placebo effect.

September 7, 2005 1:36 PM

PharmaFilter: is the placebo effect halved in a randomized trial when half the patients think there is only a 50% chance they are taking an active compound?

Assuming that this hyperthetical trial is double blind randomised and placebo controlled, and that the patients are a aware that half of them have been given a placebo... Would those sub-conscious suspicions, that I assume cause the placebo effect, be reduced as the patients rationalise that there is a good chance that just taking a sugar pill?

posted by verisimilitude to science & nature (35 comments total)

my opinion: placebo effect is not caused by sub-conscious suspicions. rather, it is a catch-all meaning what would have happened if you had done nothing. In other words, if you hadn't taken the pill. This is the biggest difficulty in measuring a medication's effectiveness--a lot of stuff goes away on its own. The reason you have to act as if you are doing something (double blind) is because otherwise the knowledge of what is really going on will affect the doctor's and the patient's observations of what is going on.
So, no, as long as its double blind, it will not reduce the placebo effect.
posted by alkupe at 1:42 PM on September 7, 2005

We only just discovered that the placebo effect is physiological. It looks like the methodology in this particular study did involve telling the subjects that they may or may not receive a pain killer. If you could dig up the actual study, it might contain more information.
posted by frykitty at 1:48 PM on September 7, 2005

alkupe: I believe the placebo effect is demonstrably more than that (what would have happened if you had done nothing). A sugar pill is more effective than no pill, even though it has no pharmacological effect on the body. It is considerably more than the fact that some stuff just goes away on its own.
posted by edd at 1:49 PM on September 7, 2005

Completely disagree with alkupe's assessment, because it has been shown that if you think you're taking a medication, it can affect how your body reacts. If you think there's only a 50% chance you've taken medication, I don't know what would happen. Interested to hear some experts weigh in.
posted by knave at 1:49 PM on September 7, 2005

My understanding of the placebo effect is actually the opposite of alkupe's. The affect is because something is being done, as opposed to nothing. Not only is there a physiological affect, the affect is present in both control and placebo groups. Patients who get face time from doctors do better.
posted by OmieWise at 1:57 PM on September 7, 2005

edd-- i believe the reason for that is because of what i outlined--knowing that you're taking medicine affects your (and your doctor's) interpretation of what your body is doing.
sorry guys, you're all wrong. nice myth, that placebo affect.
posted by alkupe at 1:59 PM on September 7, 2005

alkupe: That may be your view of what's going on, but that is definitely not the definition of the placebo effect.
posted by fvw at 2:06 PM on September 7, 2005

The placebo effect is not a myth.
posted by mfbridges at 2:09 PM on September 7, 2005

alkupe is entitled to his/her opinion, but it happens to be wrong, as s/he is confusing two phenomena: 1) The fact that many medical conditions or symptoms will go away on their own without treatment; and 2) the fact that the doctors' and patients' beliefs can influence the outcome of a treatment. Many clinical trials have both a control group who receive no therapy, and a group that receives a placebo, in order to take into account both phenomena.

First some definitions:

A placebo is a medical treatment (pill, injection, surgery, or mere hand-waving) that has no intrinsic therapeutic value as defined by Western medicine.

The placebo effect is the phenomenon in which a patient's and/or doctor's belief regarding the treatment affects its outcome. The most well-known instance of this is when a patient's belief that she is taking an effective remedy leads to an alleviation of her symptoms
despite the absence of an effective therapy.

There is considerable debate over the causes and extent of the placebo effect. One recent meta-analysis found there to be few clinical effects for conditions with objective or binary outcomes, and small benefits for conditions with subjective outcomes and the treatment of pain. Others have shown the placebo effect to be more effective, and to vary with the type and intensity of the treatment (e.g., a bigger pill results in a stronger placebo effect than a smaller one). In general, the placebo effect has been found to be strongest with regard to pain remedies - over-the-counter analgesics tend to be only a tiny bit more effective than placebos, for example.

The placebo effect is important enough that it has to be taken into account when testing new treatments. Thus the use of double-blind clinical trials, in which neither the patient nor the physician is knows whether the patient is receiving a real treatment or a placebo. Note that both the doctor and patient are in the dark, because it has been found that physicians' beliefs about the effectiveness of a treatment can influence patient outcomes as well - e.g., if your doctor knows you are taking a sugar pill, she may subtly communicate this fact to you. Double-blind trials also eliminate other forms of subjective bias that may creep in, such as physicians or researchers recording better outcomes for the patients that they know are receiving therapy.

To answer verisimilitude's question: The fact that all of the patients have been notified that there is a chance that they are receiving a placebo may indeed impact the outcome, but it is unlikely to reduce the power of the placebo effect by 50%. Ultimately it is irrelevant to the clinical trial procedure. As long as the placebo effect is consistent - and if all participants are notified of the possibility of receiving a placebo we expect that it would be - then its magnitude is unimportant. The point of a double-blind study is to eliminate subjective bias (which includes the placebo effect) and determine whether a particular therapy is more effective than a placebo or no therapy at all, and not to study the placebo effect per se.
posted by googly at 2:21 PM on September 7, 2005

i don't understand how what googly is saying differs from what i said. it sounds like almost exactly the same thing. there's no "real" placebo affect (as commonly understood) rather there is (quoting googly) a "phenomenon in which a patient's and/or doctor's belief regarding the treatment affects its outcome. "

isn't that what i said?
posted by alkupe at 2:31 PM on September 7, 2005

Alkupe, what you're talking about is the null hypothesis, i.e., that the medication being studied has no effect. I don't know what you think the common understanding of the placebo effect is, but it does have real (statistically demonstrable) effects that do not occur in patients who don't take a pill. What's not real about the effect is that its results do not indicate that the medication or treatment being studied is actually effective, but that's the whole reason we talk about it at all.
posted by keatsandyeats at 3:06 PM on September 7, 2005

Can I tack on a question? Is there a difference in the placebo effect depending upon how the placebo is delivered: needle, pill, inhaler, suppository, etc.? Does the size or color of the pill matter? The size of the needle? Etc., etc.
posted by Mo Nickels at 3:09 PM on September 7, 2005

Also, what if it's not the pill that's causing the placebo effect but the swallowing action? Or the water taken with it? Or the acupuncture-like effect of the needle?
posted by Mo Nickels at 3:11 PM on September 7, 2005

alkupe - think about it this way (and spell it this way:"effect"): "a phenomenon in which a patient's and/or doctor's belief regarding the treatment affects its outcome" would include things like the doctor saying "this will never work" every time they're in the room, or not trying as hard to help patients not receiveing the treatment; patients who know they aren't getting treatment giving up or dropping out; etc.

The placebo effect is a documented fact: if you give someone a pill they think will help them, it often will do so to some extent, even if the pill has no actual effects. Since this effect exists, it needs to be seperated from the effect of any study about the effects of a new medication.
posted by freebird at 3:15 PM on September 7, 2005

"The placebo effect is a documented fact: if you give someone a pill they think will help them, it often will do so to some extent, even if the pill has no actual effects."
I don't follow what keatsandyeats is saying.
first off, there's no such thing as documented facts, there's only ideology. but thanks, freebird.
second:
my point is that this "some extent" is basically indistinguishable from "what would have happened if you did nothing". I guess you can call if placebo effect if you want. But to me, its no different than a control group in any experiment--there's some benefit to doing nothing. the reason to give placebos is summed up well by googly--its because people are interpreting the results and are easily biased.
keatsandyeats i don't follow you.
It seems to me that the total evidence documented so far on this page is that there is some "real" placebo effect with pain medicine.
OK, what else you got, everybody?
posted by alkupe at 3:27 PM on September 7, 2005

Well, unfortunately it's really difficult to quantify "THE" placebo effect in any meaningful cross-trial way. What I mean is, as was previously mentioned above, the placebo group in a study is ultimately always compared to the trial group. In a good study, the goal is to match these two populations as well as the intervention being made and how it's delivered. Placebo effects vary a great deal from study to study, and because of the differences in patient population in each study, the differences in the illness being treated, and the differences in the actual intervention, there's no value in comparing placebo to placebo.

That said, you'll find for example that the placebo effect generally tends to be much higher for say trials of antidepressants over trials for arrhythmia medications, for example. This should make some intuitive sense based on how we think about the effect, but there's no meaningful way to quantify the difference.

In virtually all trials (though there are probably some cruel exceptions in the psych literature), for obvious ethical reasons, the subjects usually know that there's a chance they are getting placebo. However there is never a group within the same trial (as far as I know), who is told that they are definitely receiving the actual treatment when in fact they are getting a placebo instead. That would in theory, be the way to tell if that 50% chance affects the degree of placebo effect. But even if you could design a study like that, despite potential ethical problems, you would still introduce a new bias as the people performing the study would no longer be blinded to the group stratification. And even then, you still couldn't compare the numbers in that one study with numbers from similar studies on different patient populations.

I know that all sounds convoluted, but ultimately the 'value' of designing studies with placebo-controls is not the absolute placebo effect itself, but disproving the null hypothesis for the treatment group -- ie showing a statistical difference in that group's outcome relative to the placebo group.
posted by drpynchon at 3:29 PM on September 7, 2005

my point is that this "some extent" is basically indistinguishable from "what would have happened if you did nothing"

No! That's the point! You did something - you gave them a pill! They believe the pill had medicine, but it didn't! Nonetheless, they exhibit a response different from that if you gave them nothing! Why is that so difficult to follow?

first off, there's no such thing as documented facts, there's only ideology.

Never mind, I thought you were a rational person interested in real communication and trying to address this person's question.
posted by freebird at 3:41 PM on September 7, 2005

But to me, its no different than a control group in any experiment--there's some benefit to doing nothing. the reason to give placebos is summed up well by googly--its because people are interpreting the results and are easily biased.

I think the posts above have stated clearly that many experiments involve both a control (no medication) and a placebo (treatment of the same form as the medication but no medical effect), and that these two groups behave statistically differently, even in double-blind experiments where neither they nor their doctors know which group they are in (and so cannot be biased).

first off, there's no such thing as documented facts, there's only ideology.

It's hard to tell what this statement means, but in the kind of experiments being discussed here, there are (ideally) documented, reproducible, statistically significant results that are taken to approximate what the facts are. I don't really see how this kind of thing can be confused with ideology.
posted by advil at 3:43 PM on September 7, 2005

I think the posts above have stated clearly that many experiments involve both a control (no medication) and a placebo (treatment of the same form as the medication but no medical effect), and that these two groups behave statistically differently, even in double-blind experiments where neither they nor their doctors know which group they are in (and so cannot be biased).

That is actually generally not the case. The placebo in most trials *is* the control, and in fact, this is true in virtually all blinded trials by definition. If you have a group for which you are doing absolutely nothing (no placebo, even) then your study is no longer double-blinded, and you've introduced a bias. I guess in theory you could potentially design a blinded trial in which the people recording outcomes don't know about the intervention, and the study subjects are comatose for example, but that is not the norm.

And alkupe, you clearly have a fairly incomplete idea of what placebos are or why they are incorporated into studies, so save your ideology for another time.
posted by drpynchon at 3:58 PM on September 7, 2005

Clinical trials are explicitly designed to minimize the placebo effect, since what they're trying to measure is the efficacy of the treatment once placebo effect is taken out of the equation. If both groups think they are receiving the treatment, as happens in a clinical trial, then both groups are subject to the placebo effect in equal measure. (After all, the treatment may not work, which would mean both groups were effectively receiving a placebo.) The same is true, of course, regardless of how many people actually receive the drug -- even when no one does!

To measure the placebo effect, you would have to compare the outcome of patients who received no treatment and knew it with those who received a placebo but thought they were receiving the treatment. This experiment has of course been done, which is how we know the placebo effect exists. But if you simply measure how many people get better when they are given a placebo, you haven't measured the placebo effect, since you have no idea how many people would have got better without any treatment at all.
posted by kindall at 4:09 PM on September 7, 2005

Nice derail, btw.
posted by knave at 7:14 PM on September 7, 2005

I guess I'm not being clear...
I don't disagree with anything drypynchon and googly have said. In fact, I think they're saying the same thing as me in a much more thoughtful, reasoned manner. I read some other people that are throwing around all kinds of potential meanings for placebo ("your doctor saying it will never work...") and I think googly and drpynchon would agree its not like that.
AS (i think) WE ALL AGREE, the reason for taking a placebo is that you can't measure what would have happened without doing the experiment (take the meds) in a controlled manner. I thought i was clear on that. Its a control.
drpynchon, you can diss me all you want, but we're saying the same thing.
freebird where you say "No! That's the point! You did something - you gave them a pill! They believe the pill had medicine, but it didn't! Nonetheless, they exhibit a response different from that if you gave them nothing! Why is that so difficult to follow?"

You're begging the question. The fact is, as drpynchon points out, experimentally speaking, placebos are the same as doing nothing. That is, the reason we do placebos is as a control for the actual experiment. We can't experimentally measure what doing nothing would do, so we give a placebo in order to keep everyone blind. If we could keep everyone blind and give those patients nothing, we would do so.

Sorry about my statement on ideology--my point is that all scientific discovery is actually rooted in a complex web of assumptions. Its seems to have been a distraction, sorry i mentioned it. If I disagree with you, being told something is a "proven fact" isn't exactly helpful. My point is there really are no proven facts in science. Really, there aren't. So don't tell me something i dispute is a proven fact and just accept it. Or at least don't expect that to work...sorry i was expecting a more reasonable dialogue.

My overall point (i've thought it through a bit more)--i still don't see much evidence here to counter it--much of what people take for the placebo effect, this idea that "if you think its medicine it will do something good" is bunk. There may be some minor benefits to thinking you've taken pain medicine, and certainly the quality of care overall (which is really what placebos are there to maintain) can be important. But the general idea of placebo effect is way overblown.
googly, drpynchon do you really disagree?
posted by alkupe at 7:57 PM on September 7, 2005

alkupe : "The fact is, as drpynchon points out, experimentally speaking, placebos are the same as doing nothing. That is, the reason we do placebos is as a control for the actual experiment. We can't experimentally measure what doing nothing would do, so we give a placebo in order to keep everyone blind."

No. The placebo effect intends to measure how much of the improvement can be pegged to self-expectation driven endogenous modulation. Doing nothing is not the same as giving placebo, which is not the same as giving a therapeutic agent. The simplest way to explain the placebo effect is, if you believe something will help you, *and* if you believe that something is being done, then a cascade of neurochemical endorphin activity begins, elicited by the results of the two beliefs being held. Doing nothing will satisfy as a placebo, only if one is open to the belief that doing nothing in that circumstance is an effective course of treatment. However, any normal human being i.e. those who have grown up with modern sensibilities, are prejudiced that a medicine is generally required to treat maladies. Hence the necessity of administering a dummy pill that serves as the perceptual symbol of a medicine, despite no therapeutic potential inherent to it.
posted by Gyan at 8:30 PM on September 7, 2005

Note that in many fields of medicine trials are no longer commonly conducted against a placebo but against the guidelines-approved treatement or other marketed treatments.

For instance a trial on a new statin or combination therapy involving (say) a statin and a diuretic wouldn't be measured against a placebo but against one of the current gold-standard treatments.

Don't know if that's exactly relevant.
posted by mikel at 8:50 PM on September 7, 2005

No, alkupe. Placebos are NOT the same as doing nothing, and we certainly could experimentally measure doing nothing, by simply doing nothing. The placebo in an experiment is setup to account for everything in your experimental design that is done in the treatment group but not attributable to the intervention being studied itself. That is not the same as doing nothing, and leads to different outcomes than doing nothing, which is precisely why there is such a thing as a placebo effect.

I'll grant you that there was a controversial meta-analysis published in the New England Journal in 2001 that tried to suggest that placebos were in fact entirely powerless, except minimally in the case of pain. However this study was met with a flood of harsh critical responses.

There are numerous studies that document the power of placebos, so many so, that it's hard to know where to start. Consider the famous Lancet study from 1972 by Blackwell, in which medical students were told they were being given either stimulant or sedative pills. One group was given either one or two red sugar pills, or one or two blue sugar pills. Not only did the students believe the red pills were stimulants and the blue pills were sedatives. They also reported a dose response, such that two blue pills were more sedating than one, and two reds were more stimulating than one. The color and quantity of the pills alone was enough to affect subjective responses in even a dose-dependent manner.

To answer Mo Nickels question as well as give a second example of the power of placebos, there have been numerous studies on everything from hypertension to osteoarthritis that have found that the results in groups treated with placebos involving devices differed signficantly from outcomes in placebo groups treated with pills.

So whether I'm not giving you enough credit for actually appreciating the new controversy on the subject, or my assumption that you haven't studied this subject in great detail was true, the general consensus among medical researchers is that an effect attributable to placebo intervention does in fact exist, outside of the natural progression of a disease or state being studied. This effect may be relatively small or fairly large, varying a great deal from study to study, depending upon the study design, the cohort, and host of other factors.

But I do give you credit on an appropriate use of 'begging the question.'
posted by drpynchon at 9:22 PM on September 7, 2005

Alkupe: For whatever it's worth, the other people in this thread are correct.
posted by sirion at 9:40 PM on September 7, 2005

There's some oversimplification going on in this thread, as well as some undeserved vitriol.

The answer to the original poster's question is 'no'. It's surprising, but there it is.

In some anticonvulsant trials, administration of placebo reduced seizure frequency 40% over 'no treatment'. More or less, that is [because of|defined as] the placebo effect.

Actually, there are other possible reasons for this besides the placebo effect. For instance, people enrolling in trials of a new anticonvulsant may be at their wit's end, because they're in the middle of a month or two-month long cluster of seizures. The natural history of epilepsy is that, whether it's getting better or worse overall, it waxes and wanes, comes in clusters and then go. If you enroll people at the very nadir and then they wax, it looks like they're having a reduction in seizures. This kind of thing holds true for other illnesses, too, and these things let me laugh at overpompous folks who brandish their Class A Study like a whip.

But it's also pretty clear that *most* of that 40% is due to the effect of taking a placebo pill, whatever that effect is. The placebo pill usually contains a bit of sugar and some agent to keep it solid. Administering sugar and a bit of caking agent by slipping it into someone's food all secret-like does not reduce seizures or cause other beneficial effects. Handing it to them as a pill and saying "Here, swallow this, it may be a revolutionary medicine for seizure control" does do so.

Lately it was proposed that it's now unethical to use placebo trials of antiepileptic drugs (AEDs) in monotherapy, "because we have AEDs that work." The result: the placebo was thrown out in favor of a low-dose/high-dose trial model. Who do you suppose loved this? The drug companies, that's who; it's great to show a 45% reduction in the low dose group and a 55% reduction in the high dose group, but it's not so great to show a commensurate 40% reduction in the placebo group - it (rightly) makes the drug look much less useful than it otherwise looks.

To suggest that the placebo effect is physiological as if this were somehow an unusual idea is sort of silly. When something happens in or to a body, do you suppose it could be somehow unphysiological? What else was it supposed to be - magic?

More interesting are statements such as 'naloxone, a competitive antagonist at mu-opiate receptors, abolishes the placebo effect (in certain contexts)', because that statement gives insight into the neurophysiological mechanisms that make the placebo effect work.
posted by ikkyu2 at 11:59 PM on September 7, 2005

Oh, don't forget about the 'sham operations'. In one trial, to inject stem cells into the substantia nigra of folks with Parkinson disease, they actually did 'sham operations'.

That's right, they DRILLED HOLES into people's skulls, then inserted a cannula SIX INCHES DEEP into the midbrain, the very heart of the brain, just to inject a little bit of saline solution.

This is a control: it was designed to control for placebo effect as well as for the effects of every potentially active ingredient of the operation, except for stem cells.

Interestingly, the sham-operated patients improved a bit on their motor functions. The stem cell patients improved a LOT on their motor functions but their quality of life actually uniformly went down, not up, owing to disabling dyskinesias that developed as a consequence of the therapy.

That study was stopped early. What did it tell us? It told us not to inject stem cells into the nigral body. OK, I'm rambling.
posted by ikkyu2 at 12:06 AM on September 8, 2005

ikkyu2 : "The answer to the original poster's question is 'no'. It's surprising, but there it is."

Not that I'm doubting you, but can you point to the studies that demonstrate this?
posted by Gyan at 1:56 AM on September 8, 2005

To add a bit more for Mo: pill size and colour have both been shown to influence the placebo effect, as I'm sure have other similar factors.
posted by edd at 4:17 AM on September 8, 2005

I'm with Gyan on this, ikkyu2. I don't know of any studies that suggest that reduced expectation of getting a true intervention doesn't alter the placebo affect.
posted by drpynchon at 5:50 AM on September 8, 2005

Another thing to consider is active vs. passive palcebos. This is an area of controversy in anti-depressant study design, where sometimes the measured placebo effect (even after throwing out all 'extraordinary responders') comes close to the effect of the medications. If pt's are given an inert placebo, they are more likely to know that they are not receiving the medication than if they are given something that mimics at least one of the side effects, but none of the benefits, known to be associated with the medication.
posted by OmieWise at 6:14 AM on September 8, 2005

first off, there's no such thing as documented facts, there's only ideology. but thanks,

If you don't belive in science, why are you posting in the "science and nature" section?
posted by delmoi at 7:39 AM on September 8, 2005

Gyan: I didn't find the reference in the textbook I was expecting to find it it. I'll try to dig it up for you.
posted by ikkyu2 at 7:17 PM on September 8, 2005

ikkyu2, it seems that the degree of expectation does matter. As per this article Placebos and Painkillers: Is Mind as real as Matter?* in Nature Reviews Neuroscience, "In fact, in a meta-analysis of the power of placebos [87], small placebo effects were found in some clinical trials. This was probably due, among other factors, to the fact that a 50% chance of getting a placebo was openly communicated to the patients. When only the experimental placebo studies were considered, in which the information about the placebo was 'you will be given a powerful analgesic drug', larger placebo effects were observed [88]. Therefore, manipulating the degree of expectation in the laboratory setting changes the degree of the placebo effect." (Box 1)

Here's the abstract for REF 88:

A comparison of placebo effects in clinical analgesic trials versus studies of placebo analgesia.

"A previous meta-analysis of clinical analgesic trial studies showed generally low magnitudes of placebo analgesia (N. Engl. J. Med. 344 (2001) 1594). However, as studies included in their analysis used only placebo as a control condition, we conducted two meta-analyses, one in which 23 studies used only placebo as a control condition, and one in which 14 studies investigated placebo analgesic mechanisms. Magnitudes of placebo analgesic effects were much higher in the latter (mean effect size=0.95) as compared to the former (mean effect size=0.15) and were significantly different (P=0.003). This difference as well as differences in effect sizes within studies of placebo mechanisms may be parsimoniously explained by differences in expected pain levels produced by placebo suggestions and by conditioning. Furthermore, some of the studies of placebo analgesic mechanisms indicate that the magnitude of placebo analgesia is higher when the placebo analgesic effect is induced via suggestion combined with conditioning than via suggestion alone or conditioning alone. Based on these findings, we suggest that placebo analgesic effects are most optimally conceptualized in terms of perception of the placebo agent, and therefore a new definition of placebo response is proposed."

*I'm able to access the article from a home computer i.e. with no subscription. Which is strange, but it works.
posted by Gyan at 9:52 PM on September 8, 2005