Is “The” Placebo Effect Proven
A recent study, as reported in the New Scientist, purports to catch the placebo effect in the act using functional MRI scanning. This is an interesting study, and does for the first time show a neurophysiological correlate to reported placebo decreases in pain reporting.
However, reporting of the study highlights, yet again, widespread misconceptions about the nature of placebo effects – specifically that there are many placebo effects, not one placebo effect. Any reference to “the” placebo effect is therefore misleading – it is a convenient short hand, but unfortunate given prevailing misconceptions.
What most people mean when they say “the” placebo effect is a real physiological effect that derives from belief in the effects of a treatment – a mind-over-matter effect. However, the placebo effect, as it is measured in clinical trials, has a very specific operational definition. It is any and all measured effects other than a physiological response to the treatment itself.
This includes any physiological responses to belief in the treatment, but also a host of psychological factors such as reporting bias, confirmation bias, risk justification, and assessment bias. It also includes non-specific effects of being in a clinical trial – people treat themselves better when they are being observed, when they are being reminded of their illness because of frequent attention, and when they are encouraged by the hope of benefit. Such things actually affect compliance with other treatments and healthy lifestyles – in other words, people will be more compliant with other medications they may be on, and may eat better and exercise more, etc.
These variables and others are the reason for double-blinding experiments. Without doubling blinding, these placebo effects will be mixed in with the physiological effects of the treatment, if any.
Also – many people incorrectly conclude that “the” placebo effect cannot exist in small children or animals, but this is profoundly incorrect. Placebo effects result from observer bias as well – from whoever is interpreting the effects of the treatment on an animal.
It must also be pointed out that measured placebo effects differ greatly depending on the disease and the outcome being studies. The greatest effect is for pain, typically from 25-35%. This makes sense in that pain is a subjective experience and subject to a host of modifying factors, such as mood and expectation. But also, it has been known for a long time that there exists in the body natural opioids called endorphins that bind to receptors and inhibit pain, the same way the most powerful pain killers do. Therefore there is a known physiological mechanism by which mental effects could inhibit pain.
Another area where there is a strong placebo effect is any disease that is worsened from psychological stress, such as the risk of heart attacks. Any intervention, or just the act of being treated, that might reduce stress therefore has a known physiological mechanism by which disease can be mitigated.
But for other diseases, where there isn’t a known physiological mechanism, measured placebo effects are much smaller. Perhaps the most dramatic example of this is cancer survival. Here we have an aggressive disease and a very definitive outcome – death or survival. It turns out there is practically no placebo effect when it comes to cancer survival.
And here is the real problem with conflating all placebo effects as if they were one giant mind-over-matter effect – people think because there is a large and provable placebo effect for pain, there is therefore a large and provable placebo effect for everything, and that this is evidence for some mystical mental effect over the body (in the absence of known physiological mechanisms).
This brings us to the current study. What the researchers did was look at an experimental model of pain – heat applied to the skin. They then compared an analgesic cream to a placebo cream, and subjects reported about 26% less pain even with the placebo cream (right in the middle of the typical placebo response to pain). This is nothing new, but they also looked at the spinal cords of the subjects and found that the placebo pain responders had a decrease in the signal in the pain pathways similar to the response to the actual painkiller.
The press is making it seem as if “the” placebo effect is finally proven – typically missing all of the nuance of this issue. However, while interesting, this study does not really add much new to our understanding. We already knew about the placebo effect for pain, and the prevailing hypothesis is that it was due to endorphins, and therefore we would expect an actual decrease in pain signals in the spinal cord. This study confirms it – but doesn’t change our thinking about it. Also, the study did not in any way investigate the mechanism of the decrease. We assume it is from endorphins, but that’s it.
This is not to knock this study in any way – just reporting about it. It is a nice proof of concept, and opens the way to further studies to look at which brain areas are involved in placebo effects for pain. It would also be interesting to see if there is any difference in brain activity between placebo responders and non-responders. This may even lead to ways in the future of optimizing pain placebo effects, or triggering them non-pharmacologically.
I think it is important to make these distinctions regarding the placebo effect because it is so widely misunderstood and this confusion is exploited to support all sorts of unscientific and even harmful medical modalities and interventions.
I think this article is pretty relevant when looking at “the” placebo effect.
http://content.nejm.org/cgi/content/short/344/21/1594
Basically they compared results in studies where there was both a placebo group and a no treatment group (ie. some people were given an actual drug, some were told they were getting a drug but got a placebo, and some just got nothing).
They found that only in instances of scaled subjective measurement (ie. rate the amount of pain relief out of 100) was there any significant difference between placebos and no treatment. They further found that this effect was reduced the larger the sample size.
They concluded that there was little evidence that any kind of significant “placebo effect” actually existed.
Numerous articles have been published on the “placebo effect” even 80 years ago. An awful lot (data from tests) is known, and yet very little is understood. This is why the working definition in what Steve Novella mentions above: the placebo is anything which cannot be ascribed to the treatment being tested. So in the article you mention, sure in those circumstances there may be no difference between placebo and nothing for large populations, but there are other (even a few large) placebo vs. nothing experiments which show that a placebo is better than nothing. It really depends an awful lot on what you’re testing (usually managing pain of some sort).
It would be fantastic if people can eventually identify mechanisms associated with specific types of placebo effect for specific ailments (in placebo vs. no treatment trials) because that might suggest an appropriate target for drugs which can be used in conjunction with existing treatments. However, the study mentioned above shows a nice correlation – but doesn’t even convince me that it is anything other than a correlation, nor does it even suggest to me a further line of investigation which might open up new possibilities. In short, an experiment has been done – but does it answer any questions and does it expose new questions? If not, then the experiment and its results are of no value.
If we can’t use the term “the placebo effect” to describe the phenomenon of the ‘mind over matter’ effect, where the act of thinking a treatment will help, causes it to work, then what term should be used?
I’ve understood it that this is the placebo effect, and the other phenomena you described that the placebo group controls for are just other confounding factors. In other words, the placebo group controls for the placebo effect, plus other effects that can occur.
Here you refer to confounding factors as placebo effects yet according to
Bausell’s “Snake Oil Science” and Singh/Ernst’s “Trick or Treatment” the mind-over-matter type of confounding factor is “the placebo effect”. Bausell even talks about a study where researches were able to block “the placebo effect” using opioid blockers. He had clearly separated the placebo effect from other confounding factors. I apologize if my terminology is off, I’m not a doctor/scientist.
I realize this is a debate about semantics (the definition of placebo), but this is an extremely confusing area therefore consistent and clear terminology would help immensely.
I’m also curious if all the doctors in the skeptical community are in agreement on its definition. You can see here some debate about what the placebo is on sciencebasedmedicine.org (look in the comments).
The placebo is a lot more than just believing something might work, and it is all encompassed in the definition Steve gives in his post. Also keep in mind that that is the operational definition when you actually have an untreated control group; in many cases there is no such group because there is a currently accepted treatment to compare against and you’re searching for something better in some respect (fewer allergic reactions, higher efficacy, etc). The definition does vary a lot if you’re interested in studying placebo effects as such – but those details are absolutely meaningless in the case of trials of new drugs or new treatments.
Pain management is where you’ll see a lot of discussion of placebo effects because there are many types of pains for which there simply aren’t any great drugs to use, so anything which can help the patients feel better is a good thing; this broad category of placebo effects is often *incorrectly* touted as “mind over matter” – though the phrase is used often, it is meaningless – it does not help us understand anything and it explains nothing. But even in pain management, placebo effects are necessarily limited to the specific origin, intensity, and perhaps even location of the pain addressed in the study. To do a small study such as the one mentioned and claim that the placebo effect is somehow “proven” is utterly ridiculous and useless.
Statistics and studies aside, it seems to me that the `happier’ people are the easier it is for them to deal with physical problems. Of course a placebo has no specific medical effect and can’t replace treatment, but a placebo can be as little as a kiss on the cheek from a loved one (endorphins galore!) and lighten some painful moments. Or is this just an old woman talking?
There is no one consistent definition for the placebo effect, so it is used in various contexts and this causes confusion.
I have read more careful texts that refer to the “analgesic placebo effect” or “placebo effect for pain” to be more specific.
Keep in mind, as I wrote – there is an operational definition for “placebo” in clinical trials. This is -the effect measured in the placebo arm of the study. That is considered as the placebo effect for that study, and the effect of the treatment is the difference between the treatment arm and the placebo arm. That “placebo effect” is, as I said, everything but a physiological response to the treatment. It contains confounders and non-specific effects.
The problem comes from equating the measured placebo effect in a clinical trial (everything but the treatment effect) with the mind-over-matter placebo effect. And also from equating placebo effects for subjective symptoms, like pain, especially where there is a known mechanism of mental modification – with disease outcomes where there is essentially no placebo effect other than confounding factors.
“specifically that there many placebo effects, not one placebo effect.”
Should be “specifically that there are many placebo effects”.
Good article though.
Tangentially related to placebo effects, I listened to last weeks show and was alarmed to hear that the NIH somehow got muscled into including acupuncture as a valid treatment.
Alarmed? Me, too, as well as openly amazed when my regular GP doc, during a recent check-up, said he approved of acupuncture as a treatment…
Individuals with different worldviews must respond to different placebos. Prayer would not work for me, but I’m sure it would for my sister. I have friends for whom acupuncture would be worth a try, but probably not for me; I would need a pill or shot that I believed was real medication. Have any studies been done to investigate this kind of difference?
Reminds me of the argument that you really have to believe in the woo for it to work.
“Placebo effects result from observer bias as well – from whoever is interpreting the effects of the treatment on an animal.”
Curious, I had never considered this, even though I work/volunteer for a rescue/sanctuary organisation for abused/neglected aves.
Nice piece Dr. Novella, interesting information, well done.
Steve,
I searched the Science-based Medicine blog for “Association of Health Care Journalists” and “AHCJ” and didn’t find anything. What’s their contribution to medical reporting?
Is anyone here actually in research? If so, I would like to make a suggestion based on ‘Placebo Effect’ observations I have read about. This goes to the additional research by Prof. Fabrizio Benedetti of the University of Turin, his addition to the Placebo Effect…
http://hal9000.cisi.unito.it/wf/DIPARTIMEN/Neuroscien/Fisiologia/Staff/Professori/EmbeddedText.htm_cvt.htm
The naloxone, (morphine effect blocker,) should not have effected the Saline, but it did. We know the effect was not ‘the mind’ because it was in error. The mind knowing the effect of the naloxone would have done nothing because the lack of morphine, a requirement of the naloxone.
Consider that the administering person knew… Biochemical communication, possibly in Pheromones or Hormones would explain the patients mind knowing what to do. Once an extracellular ‘first messenger’ binds its receptor to a cell surface, a signal is transmitted to the cell interior, producing a biological response.
Simple research here. Take an oncology center and charge the rooms with successful patients, (hire people that were cured,) and check the rate of success… If it increases, we can assume the cancer survivors transmitted the cure.
This could explain the placebo effect working across multiple platforms. Biochemical communication takes away the mystery as well if it works, because we do communicate that way. Our arrogance assumes it is our mind…
As a note, the saline worked as morphine due to the positive reinforcement of three days with the real thing.