Articles

Michael Specter on the Placebo Effect

Michael Specter is a good science journalist. I particularly enjoyed his book, Denialism. In a recent New Yorker article he tackles the difficult question of the placebo effect in modern medicine. While he does a fair job of hitting upon the key points of this question, I think he missed some important aspects of this question and allowed the views of Ted Kaptchuk to overly influence the framing of the article. Specter fell for the typical journalist trap — frame the article around a charismatic “maverick”, complete with compelling anecdotes, bury the meat of legitimate skepticism deep in the article, but then bring it all back to the maverick in the end. Be sure to tell us how this is going to change everything. This is good story telling, but very problematic as science journalism.

Kaptchuk himself is an interesting character. He is heading Harvard’s Program in Placebo Studies and the Therapeutic Encounter. He has produced some good science on the placebo effect, but does not seem to want to draw the appropriate lessons from that research, and passes his bias on to Specter.  From the article the quotes from Kaptchuk that most strike me are those about his personal experience with placebo medicine. Specter reports:

“There was no fucking way needles or herbs did anything for that woman’s ovaries,” he told me, still looking mystified, thirty-five years later. “It had to be some kind of placebo, but I had never given the idea of a placebo effect much attention. I had great respect for shamans—and I still do. I have always believed there is an important component of medicine that involves suggestion, ritual, and belief—all ideas that make scientists scream. Still, I asked myself, Could I have cured her? How? I mean, what could possibly have been the mechanism?”

To me this is no different than Ray Hyman being flabbergasted when his cook-book palm readings were found to be highly accurate by his clients. Hyman, however, took a step back and did what a scientist should do. He gave the opposite readings as was dictated by the reference book and found that his clients were just as pleased with the results. He concluded that the accuracy of palm reading (any cold reading, really) is just an illusion, one happily constructed by the client.

It is difficult to shake the sense, however, that something real is going on in such situations. We are struck by coincidences and apparent patterns and effect. The patterns may seem so real to us that, emotionally, we just can’t accept that they are illusory.

That is what I see in Kaptchuk’s memory of that one case, which is just representative of the many cases that apparently convinced him of the power of the placebo. All practitioners, like all psychics (regardless of how sincere they are), are subject to this illusion. We are all subject to confirmation bias, statistical effects like regression to the mean, expectation bias, and reporting bias. These psychological effects create the illusion of a real effect where none exists.

In fact, the analogy to psychics is very illuminating, because it demonstrates that people can be led to believe that there is a real effect when we know (as much as we can know anything in science) that none exists. Why should the placebo effect get any different treatment?

The real question is: how much of the placebo effect is due to illusions, like confirmation bias and regression to the mean, and how much is due to a real physiological response to the therapeutic interaction? This is an important question, and Kaptchuk is right to research it, as long as he is willing to listen to the answer.

We already have much research that addresses this question, and Specter references some of it in his article. In fact, Specter would have been better served to frame his article around Asbjörn Hróbjartsson and his research, rather than Kaptchuk (the same facts would then have been framed very differently, it seems). Hróbjartsson has been reviewing studies in which a placebo arm is compared to a no-treatment group. He first published his findings in 2001, but has updated his research. His most recent study concludes:

We did not find that placebo interventions have important clinical effects in general. However, in certain settings placebo interventions can influence patient-reported outcomes, especially pain and nausea, though it is difficult to distinguish patient-reported effects of placebo from biased reporting. The effect on pain varied, even among trials with low risk of bias, from negligible to clinically important. Variations in the effect of placebo were partly explained by variations in how trials were conducted and how patients were informed.

In other words, the best research we have strongly suggests that placebo effects are illusions, not real physiological effects. The possible exception to this are the subjective symptoms of pain and nausea, where the placebo effects are highly variable and may be due to subjective reporting. In a separate publication Hróbjartsson explores this question further and concludes that the existing research is consistent with reporting bias (i.e. illusion), but we need to creatively design studies to better answer this question.

Placebo effects are mostly just as much an illusion as precognition or talking with the dead. Pain is the notable exception, which makes physiological sense. Pain is a subjective experience, evolved to have adaptive features that are highly situational. There are times when pain should be very bothersome, and other times when it’s more adaptive to be able to ignore pain. So it is no surprise that mood and expectation can highly influence the reporting of pain.

There is another way to get at the question of how much of placebo effects is real and how much is illusion, and ironically Kaptchuk himself has authored one of the best studies of this kind — comparing subjective and objective outcomes. Kaptchuk performed the now famous (at least among medical skeptics) study of medical treatment vs placebo vs no treatment of asthma patients. David Gorski reviews the study here, pointing out that for subjective outcomes there was a measurable placebo effect. For objective outcomes, there was none. That is the key finding. In my opinion it is difficult to miss the implications of this — that at least when it comes to asthma the placebo effect is illusion, not real.

Kaptchuk, however, missed that. He concluded, rather (to quote Specter):

Kaptchuk concluded that objective data should not be the only criterion for doctors to consider. “Even though objective physiological measures are important,” he wrote in the study, published earlier this year in The New England Journal of Medicine, “other outcomes such as emergency room visits and quality-of-life metrics may be more clinically relevant to patients and physicians.”

I wonder what he would have concluded if the study showed there was an objective improvement with placebo? Anyway — Kaptchuk just wants to change the rules to suit the outcome, rather than listen to the outcome. Asthma is perhaps the perfect disease to demonstrate the folly of his approach. Asthma attacks can kill. If someone is made to feel that they are better because of expectation, investment justification, and other psychological effects — essentially biasing their reporting of their symptoms — while the physiological asthma attacks continues, they are theoretically at higher risk of complications, including death. Kaptchuk, however, pulls the standard — what are you going to listen to, numbers or people. He misses the point. The subjects of his study who received the placebo were not better. They just reported that they were because they felt they were expected to.

If the study were just a pain study, Kaptchuk might have a point. As I teach my students every day — you have to ask what the goal of treatment is. If it is solely to make the patient feel better, then it’s OK to assess a treatment (in clinical practice, not as a scientific study of efficacy) on whether or not the patient feels better. But when your goal is to change the course of a disease, or prevent a complication, then you have to use objective measures, not subjective report.

Conclusion

There is no question that if you give someone a treatment and ask them how they feel, they are likely to report that they feel better. There are numerous psychological ways to manipulate and maximize this response (type of treatment, cost of treatment, and the interaction of the practitioner). No one doubts this. The real question we have to ask, however, is — is the reported improvement the same type of illusion that makes people feel psychics have magical power, or is it a real physiological effect that can be exploited?

The best evidence we have strongly suggests that the answer (for everything but pain and nausea) is that the placebo effect is largely or entirely an illusion. It is something to be controlled for so that we do not fool ourselves into thinking that a worthless treatment has a real effect. For pain and nausea the question is a bit more complex, but still the best evidence suggests that any placebo effect is small and variable, and may also be largely biased reporting. To the extent that placebo effects for pain are real, we can get them by having a pleasant therapeutic interaction.

It is unlikely that spending billions of dollars on further research is going to give us the ability to exploit the placebo effect for pain any better than we can now, simply by being hopeful and supportive of the patient. That will not stop billions of dollars from being spent on asking and answering the same questions over and over again.

Specter missed, in my opinion, a major context to this question. Proponents of so-called CAM (complementary or alternative medicine) have been telling the world for decades that their treatments are real, and they can prove it once they are studied properly. Now that they have been (at least all the major CAM modalities) and found not to work any better than placebo, CAM advocates have come out with a range of special pleading, such as the bogus claim that “Western” science cannot study such mystical methods.

Recently the main special pleading we have been hearing is that CAM modalities work through the placebo effect. That is why they are so desperate to represent the placebo effect as something more than it is — as a real effect worthy of exploitation. If the placebo effect is all illusion, however, then CAM is all illusion.

Posted in: Science and Medicine

Leave a Comment (21) ↓

21 thoughts on “Michael Specter on the Placebo Effect

  1. starskeptic says:

    Yeah, there’s a “placebo effect” like Atheism is a “god effect”….

  2. cervantes says:

    FYI, Kaptchuk and the gang are spending millions of dollars of NIH money — i.e, your money — to put people receiving placebo treatments (specifically acupuncture) in fMRI machines and see how their brains light up. The idea is that they thereby demonstrate something “real” is happening. Of course our brains respond to experience — and? This whole boondoggle really frosts my pumpkin.

  3. Harriet Hall says:

    Classifying placebo as illusion doesn’t mean we can dismiss it, and I don’t think that’s what Dr. Novella is saying. According to research by Benedetti and others, these illusions themselves can sometimes cause real physiologic effects. If we can better understand what happens in the patient’s brain and body we might be able to figure out how best to ethically use those processes for objective benefit or to enhance patient comfort. The lesson of the asthma study is that placebos can harm: the illusions can be misinterpreted as real improvement and interfere with effective treatment.

    I recently read a compelling analogy with counterfeit money: it is effective for buying things and therefore benefits those who are given it and use it, but it’s not real. People who unknowingly receive it are deceived, and when they use it without understanding what it is, it can lead to unfortunate consequences.

  4. nybgrus says:

    this all make sense – and I was intensely interested in Benedetti’s work. The take home message for me is that there are indeed actual benefits to placebo – both physiologically and psychologically. These are not to be dismissed or taken lightly.

    However, we know the physiological effects are very limited and have significantly varying response rates. We also know the same is true for psychological effects. The precise extent of this is, of course, yet to be definitively delineated.

    We also know that people are complex and individual variability in both physiology and psychology of a person (especially a person as a patient) is the rule, not the exception.

    This all leads to the conclusion that for clinical practice, placebo effects are something to be aware of so as 1) to maximize them for both the therapeutic effect and psychological well being of a patient and 2) to not be fooled into thinking an actual response has occured for a condition where modifying the disease course is integral to outcomes.

    However, it further elucidates that for any condition (and only in cases of chronic pain can an argument even be made, however it fails, IMO) utilizing a treatment that is pure placebo is simply ineffective and likely dangerous.

    In other words, placebo effects are a very useful adjunct to proven therapy. However, when there is no proven therapy, placebo effects cannot be used instead.

  5. David Gorski says:

    He has produced some good science on the placebo effect, but does not seem to want to draw the appropriate lessons from that research, and passes his bias onto Specter.

    Actually, part of good science is drawing appropriate conclusions from the data. If you don’t draw appropriate conclusions from the data, in my mind you aren’t doing good science. :-)

  6. William M. London says:

    Thank you, Dr. Novella, for another important response to placebo hype!

    You wrote: “In a separate publication Hróbjartsson explores this question further and concludes that the existing research is consistent with reporting bias (i.e. illusion), but we need to creatively design studies to better answer this question.” I’m surprised you didn’t mention in that sentence that one of Hróbjartsson’s co-authors of that publication is Kaptchuk.

    You did not entirely rule out pain and nausea as sometimes being altered by true placebo effects (as opposed to those reflecting response bias and other bias). I suspect there might be other manifestations of illness truly affected at least in the short-term by the process of administering treatment and not simply the treatment itself. The process of administering treatment may in some instances help patients relax and thereby turn down physiological stress reactions that can exacerbate health problems and lead people to seek treatment.

    A theme in the literature on psychedelic drugs is that personal perceptions and physiological responses can be very pleasant or very unpleasant depending upon psychological set and social/physical environmental setting. That notion is similar to the notion that good bedside manner might in some instances be at least as therapeutic as the supposedly active treatment.

    The literature on balanced placebo designs for studying reactions to alcohol (e.g., sexual arousal) suggests that there are real physiological and behavioral effects of expectations. Perhaps expectations matter for some physiological reactions to some kinds of medical treatment as well.

    I wonder what you think about the evidence for any kind of true nocebo effects and the role of Pavlovian conditioning, in particular, in eliciting them. It seems likely that nocebo effects are more robust than placebo effects largely because it’s easier to harm than it is to help.

  7. pmoran says:

    The only illusions we are sure of are those giving rise to beliefs in intrinsic treatment efficacy.

    They go on in the doctor’s mind.

    We don’t yet know for sure what is going on in the patient’s mind (or body, entirely, for that matter). It is not even easy to place upper limits upon potential subjective benefits, since Hrobjartsson style studies using untreated and placebo groups are likely to have designs that inhibit placebo influences, or at least not evoke them to the fullest.

    Lets try to separate the science from the urge to demonise CAM.

  8. Harriet Hall says:

    @pmoran,
    “Hrobjartsson style studies using untreated and placebo groups are likely to have designs that inhibit placebo influences, or at least not evoke them to the fullest.”

    Please explain what you mean. It sounds like an insult to Hrobjartsson to say he is not designing studies capable of testing what he is trying to test. It almost sounds like special pleading that what you want to believe can’t be properly tested.

  9. ConspicuousCarl says:

    David Gorski on 07 Dec 2011 at 2:04 pm
    Actually, part of good science is drawing appropriate conclusions from the data. If you don’t draw appropriate conclusions from the data, in my mind you aren’t doing good science.

    This brings up another issue which applies not only to Kaptchuk, but to a lot of other people on the fringes of medicine. Even if the proposed subject is a valid area for scientific study (as is the case with placebos), the person making the proposal (such as Kaptchuk) might not be qualified to carry out that research if they are prone to believing in absurd and/or unsupported extrapolations.

  10. pmoran says:

    @pmoran,
    “Hrobjartsson style studies using untreated and placebo groups are likely to have designs that inhibit placebo influences, or at least not evoke them to the fullest.”

    Please explain what you mean. It sounds like an insult to Hrobjartsson to say he is not designing studies capable of testing what he is trying to test. It almost sounds like special pleading that what you want to believe can’t be properly tested.

    Not in the least. These are standard RCTs that happen to include a “no treatment” arm as well as a placebo arm.

    If properly informed, the patients in such studies are in a state of uncertainty as to whether they are getting an active treatment or not. That will surely limit placebo responsiveness.

    This helps explain why placebo responses seem robust whenever studies are designed to specifically elicit them, but less so within this kind of routine clinical research. It also helps explain why H. finds such variable results from one study to another.

    The problem remains, however, as Dr Novella says, in being able to distinguish true placebo responses from reporting biases, which is indeed something that is difficult to “properly test” with the usual kind of RCT. The bias is also liable to increase along with any true placebo responses.

    We would seem to seem to need an objective surrogate for the patient’s subjective experience.

  11. ConspicuousCarl says:

    William M. London on 07 Dec 2011 at 2:37 pm

    The literature on balanced placebo designs for studying reactions to alcohol (e.g., sexual arousal) suggests that there are real physiological and behavioral effects of expectations.

    Is there a better example than that? Showing that sexual arousal is influenced by mental states is hardly ground-breaking. In fact, I would think that it would be more interesting to find things other than expectations and subjective experiences which affect sexual arousal.

  12. Harriet Hall says:

    @pmoran,
    “If properly informed, the patients in such studies are in a state of uncertainty as to whether they are getting an active treatment or not. That will surely limit placebo responsiveness.”

    Benedetti has found a way around this objection. See the explanation of his “placebo balanced design” at http://www.sciencebasedmedicine.org/index.php/benedetti-on-placebos/

  13. ConspicuousCarl says:

    Harriet Hall on 07 Dec 2011 at 8:06 pm

    @pmoran,
    “If properly informed, the patients in such studies are in a state of uncertainty as to whether they are getting an active treatment or not. That will surely limit placebo responsiveness.”

    Benedetti has found a way around this objection. See the explanation of his “placebo balanced design” at http://www.sciencebasedmedicine.org/index.php/benedetti-on-placebos/

    Benedetti’s quadratic design sounds like a great idea, but PMoran’s quoted sentence suggests that patients be told that they don’t know (each of Benedetti’s 4 groups is told one thing or another).

    My reaction to that sentence was to think of how horrible people are at maintaining a neutral position, and how we tend to form conclusions on weak or irrelevant information. I wouldn’t be surprised if that instruction resulted in weaker effects than active deception, but I also don’t think we can assume that telling people their status is uncertain will actually result in them being uncertain.

    Shouldn’t there be a third set of 2 groups (test and control) which is told that their assignment is unknown? I would suspect that their results would be influenced more by the physical procedure (i.e., routinely clinical vs. designed deception as pmoran said) than the verbal declarations, but that might just be the armchair magician in me wanting to believe that.

  14. gmccardle says:

    In my humble opinion, the placebo effect should be lumped together with quack practices such as homeopathy, reiki and the like. The main difference is that the medical practitioner prescribing the placebo is doing so with the full knowledge that it should have no physiological effect on the patient. Ok, some quacks that are honest with themselves realize that same thing. That being said, if a patient says that he/she feels subjectively better after taking a placebo does that justify its use in place of an actual therapeutic agent? I say no. Since the placebo effect is a complex psychological reaction involving many variables, there is a good chance it may work in one setting and not another. Pharmaceutical intervention, unless it is contraindicated, should have more predictable outcomes.

    You all might be interested in checking out my new skeptically themed website/podcast at The Inconvenient Truth.

  15. pmoran says:

    Yes, I am not sure how Benedetti’s design will answer the specific questions raised here i.e. how strong can placebo responses be under conditions favorable for them, and how much of the reported benefits are due to reporting biases?

    I wonder if his design arose from the old idea that placebo responses are a kind of constant within medical practice. We now know that they are very sensitive to every detail of the therapeutic interaction.

    We would thus expect placebo responses to be titratable up or down in ALL Benedetti’s groups depending upon precisely what they are told and any conscious or unconscious cueing by the investigators. We know people will respond to placebo if told that some people respond favorably to it.

  16. erastotle says:

    Do we know if psychiatric drugs have effects beyond placebo? If so, how?

    Since most psychiatric symptoms are self-reported, is this a case where placebo is a useful treatment?

  17. nybgrus says:

    After decades of research, scientists discover that the optimal administration of morphine for chronic pain syndromes requires an Adam Sandler movie to be playing (his older stuff, not the new garbage) and a beautiful nurse of the appropriate sex and orientation for the patient pushing the drug whilst standing on only the right foot and humming Steve Miller (preferably Joker). If the syringe can be inserted into the luer lock at the culmination of a triple Salchow that is preferable, but scientists realize not everyone can perform such a feat and thus we must resign ourselves to having most morphine administrations lacking the scientifically optimized delivery of placebo effects.

    Researchers are continuing to work hard to tease out the precise method for maximizing placebo response in vaccine and antibiotic administration. They are optimistic that once perfected, the drugs themselves may become obsolete. Preliminary research demonstrate that wearing a fake mustache may be the key to cracking this nut.

  18. William M. London says:

    pmoran,

    The balanced placebo design isn’t uniquely Benedetti’s design. It’s old news. G. Alan Marlatt used it in the 1970s. There are methodological and ethical challenges in implementing persuasive placebo conditions and manipulating expectancies of research participants, but I can’t think of a better way to distinguish expectancy effects from pharmacological effects. That’s why it has been used for decades by alcohol researchers and continues to be refined.

    Perhaps the best ways to separate participant reporting bias from what might be called true placebo effects are (1) to focus on physiological endpoints (which may be responsive to Pavlovian conditioning or conscious expectations) and/or (2) pay close attention to the demand characteristics of interactions with patients when outcomes are self-reports. See http://en.wikipedia.org/wiki/Demand_characteristics

    When outcome variables are subjective, it is difficult to rule out reporting biases and hard to tell how much is left over for real relief of significant magnitude. We know that reporting biases are real from a large body of social psychology investigation. There is much wisdom in the message: “There are no greater liars than quacks, except for their patients” (attributed to Ben Franklin).

  19. pmoran says:

    Billy London: When outcome variables are subjective, it is difficult to rule out reporting biases and hard to tell how much is left over for real relief of significant magnitude. We know that reporting biases are real from a large body of social psychology investigation. There is much wisdom in the message: “There are no greater liars than quacks, except for their patients” (attributed to Ben Franklin).

    Agreed. Yet there is at least equal plausibility to medically significant placebo responses.

    - Potential mechanisms abound, at their simplest through merely paying less attention to symptoms having taken action concerning them and getting on with normal life.

    - Mankind seems highly responsive to suggestion in other respects, why not in medicine?

    - We doctors are only too willing to incriminate psychological factors in the production of some kinds of illness — why not in the relief thereof?

    - Placebo responses would have survival benefits if evolving humanoids “learnt” to respond to random herbs or primitive religious rites when impaired by illness.

    - The evidence for significant placebo responses is even firming up a little, as preliminary evidence emerges of neurophysiological and other (very limited) physiological effects from placebo.

    - With regard to CAM, placebo responses help explain some otherwise tantalizing testimonials, indeed why CAM can thrive at all in present times. Observe the euphoria, the sense of well-being, that CAM users often express having completed one of their scientifically ridiculous ” cleanses”. These may be very temporary emotional reactions, but they are likely vivid enough to help some people cope with, or get over, or reevaluate some kinds of illness .

    Yes, patients can lie. We all lie out of politeness and sometimes for other kind reasons — “Mummy will kiss it better.” But Ben Franklin would have made his pronouncement before there was any general appreciation of the possibility that people might really feel better with quack cures because of placebo responses.

  20. Nikola says:

    @nybgrus

    Awesome.

Comments are closed.