The Power of the Placebo: a skeptical view


Something has always bothered me about the placebo effect. And I don’t just mean the way it is co-opted by advocates of pseudomedical bollocks to justify their claims. I’ve lost count of the number of people I’ve heard defend homeopathy or acupuncture with the statement “It does work – it just works through the placebo effect”.

Well, no. If something works as well as a placebo, that’s the same as saying it doesn’t work. If a plain sugar pill works as well as a homeopathically treated sugar pill then – whatever else might be helping the patient feel better – it is not the homeopathy.

No, my problem is why a placebo helps at all. How can actively doing nothing be different from passively doing nothing? Mind over matter? I’m skeptical.

In 2001 the New England Journal of Medicine published a paper titled “Is the Placebo Powerless – An Analysis of Clinical Trials Comparing Placebo with No Treatment”. This was a systematic review which looked at clinical trials featuring  both a “placebo” and a “no treatment” wing.

The researchers looked at 130 trials, sixteen of which were excluded for lacking the relevant data. Of the remainder, thirty-two trials had binary outcomes (as in, yes it worked, or no it didn’t) and eighty-two had continuous outcomes.

The study concluded that placebos had no effect on binary outcomes. That is to say, when the results were a simple yes/no, black/white, did it work vs did it not, then the placebo performed just as well as no treatment. Which makes sense, of course, as both placebo and nothing do, er, nothing. It’s just that when you take a placebo you get to pretend otherwise.

For objective, continuous outcomes the results were much the same. When we are able to objectively measure the impact of an intervention, like “has the swelling gone down?”, or “is the bone healing?”, then again placebo performs as well as no treatment.

But subjective continuous outcomes were different. Subjective outcomes are outcomes which involve some judgement by the patient or the clinician. “Do you feel better?” or “Are you less nauseous?” We are unable to objectively test how nauseous the patient feels, so we have no choice but to ask them how they are feeling.

The paper concludes:

Although placebos had no significant effects on objective or binary outcomes, they had possible small benefits in studies with continuous subjective outcomes and for the treatment of pain.

So let’s say we have a patient with a grazed elbow. Pain receptors around the graze will fire, sending nerve impulses up her arm, up her spine, and into her brain, which interprets the impulses as elbow pain. There is a strong psychological component to how she experiences that pain, and to that extent I can readily accept that her expectations – as managed by a placebo – could modify her experience of that pain.

Although the pain receptors are still firing at the same rate, and the signals are still travelling up her spine to her brain, the fact that she has been told the sugar pill she was given is actually a powerful painkiller means that she reports, and may even experience, less pain. Even when the nerve signals arriving at her brain are unchanged.

So far, so good. Nothing here which offends my reductionist sensibilities. But then came Ben Goldacre and Bad Science, a book which should probably be put on the National Curriculum. It’s a magnificent work, though the chapter on the placebo effect left me scratching my head.

In Bad Science, and in many dozens of lectures since, Goldacre enthuses about the amazing power of the placebo. By cunningly comparing one placebo to another, he says, we can discover a lot about the size of placebo effects. For example, we can tell that two placebo sugar pills are better than one sugar pill. We know that large sugar pills are better than small sugar pills, but tiny sugar pills are better than both. We can confidently make statements like pink sugar pills help you concentrate better than blue sugar pills do, even though they are both pharmacologically inert.

But that’s crazy!

Most bizarrely, to me at least, was his description of a study by the anthropologist Daniel Moerman. Ingeniously, Moerman compared the placebo wings of several gastric ulcer trials to see how the relative placebo effects compared. Gastric ulcers are good, because you can stick a camera down the patient’s throat to see if the ulcers have healed. This is a nice, objective outcome – something that the New England Journal of Medicine tells us should be immune to placebo effects.

But Moerman’s study says different. Moerman found that taking four placebo pills makes gastric ulcers heal faster than taking two sugar pills. Let me reiterate that for emphasis: gastric ulcers physically heal and disappear from your gut faster when you take four inert pills, versus two inert pills. That’s not a psychological overlay on a continuous subjective anything, that’s a solid, objective outcome. And, as Goldacre says, is an outrageous finding.

Is this some kind of mind over matter? Not in the woo-woo sense, of course, but does this show us that the brain can actually direct the body to heal a gastric ulcer faster some how, because it thinks it has taken medicine to make that happen? Last time I checked, 4 × 0 and 2 × 0 are both zero, so how can four inert pills be different to two?  What is the physiological mechanism governing this? What is the brain doing to provoke a physical different in how ulcers heal?

I don’t understand. How can this work?

To my rescue came Steve Novella and the Science-Based Medicine blog.   Last month, the Wall Street Journal published an unfortunately uncritical article about acupuncture, which repeated some crazy claims about how both acupuncture and placebo prompt the body into healing itself, which is why acupuncture and placebo produce similar results in clinical trials.

“I don’t see any disconnect between how acupuncture works and how a placebo works,” says radiologist Vitaly Napadow at the Martinos center. “The body knows how to heal itself. That’s what a placebo does, too.”

This prompted Novella to write an article called Placebo Effects Revisited, which featured an interesting characterisation of the placebo effect, as observed in the context of a clinical trial.

“Placebo effects, as measured in clinical trials, includes a host of factors – everything other than a physiological response to an active treatment.

“These placebo effects include the bias of the researchers, the desire of the subjects to please the researchers and to get well, non-specific effects of receiving medical intervention and attention, and other artifacts of the research process.”

This means that placebo data published in clinical trials includes all sorts of stuff, not just the physiological effects of inert interventions.  Novella goes even further, suggesting that placebo effects don’t just include things like researcher bias, but are mostly comprised of bias in reporting and observation and non-specific effects.

In this context, it suddenly looks to me like Daniel Moerman, and perhaps Ben Goldacre, are comparing apples to oranges.

When you compare the placebo wings of different trials, even for the same illness, you’re doing more than comparing the relative effects of various inert treatments. You’re also comparing biases of the researchers involved, will which differ across trials.

Could it be that the reason four pills make gastric ulcers heal faster than two pills is not because of some outrageous mind-over-matter effect, but is simply due to researcher bias, or other artifacts which exist only in the data and not in the patient?

I don’t know. I’m not a doctor, nor am I remotely qualified to draw a conclusion.

But I would be very interested in the results if someone who knew what they were doing actually did the research to find out.

, , , ,

  1. #1 by Doc Bill on April 20, 2010 - 13:13

    Fascinating topic and a great review which has made me think at lots of levels. Clearly, when the end point is subjective, then the magnitude of a placebo effect can be affected by a whole raft of determinants, both inside and outside the patient. What about objective measurements like time to healing? Well, we know that mood has an effect upon the autonomic nervous system. If you are waiting for an interview or examination then all sorts of symptoms are common, sweating, dry mouth, abdominal pain etc, which are mediated via the autonomic nervous sytem. We know that gastric acid secretion and gut motility are controlled by the sympathetic and parasympathetic nervous system, so that if someones mood and beliefs are affected by a belief in a treatment, then there is a clear mechanism by which physiological events can occur. Cut down acid secretion and some ulcers will heal.

    Im wittering a bit, but I’m a committed evidence based medicine doc and skeptic, but lets remember that absence of evidence is not the same as evidence of absence of an effect. Sometimes we’re not asking the right questions, sometimes we’re not going about finding the answers in the right way and sometimes its all just too hard!

  2. #2 by Jon d on April 20, 2010 - 23:52

    On the subject of comparing apples and oranges you’re looking at a medical paper (Hróbjartsson & Gøtzsche) in a medical journal and an anthropological paper (Moerman) from an anthropological journal. It’s possible that it’s both valid for anthropologists to scratch their chin over Moerman’s findings and for medics to not admit them to their canon above the status of anacdote.

    Moerman found that taking four placebo pills makes gastric ulcers heal faster than taking two

    I can’t see the full text of Moerman btw, Does he really say 4 placebo’s *make* ulcers heal quicker than 2 or does he say something less emphatic about correlation?
    (Correlation famously does not imply causation)

  3. #3 by Jon d on April 21, 2010 - 11:16

    Something else occurred to me last night. You can write it up as if the doctors shove the gastroscope down the patients throat and objectively measure the ulcers without so much as a ‘by your leave’ to their experimental subjects but it occurs to me that probably they gave it a bit of the ‘how are we feeling today Mr Jones? first like you’d expect in any other sort of examination.
    *Even if the 2 and 4 pill trials were identical in every other respect* we still know from trials designed to investigate the placebo effect that placebo pills from flashy boxes are more effective at reducing subjective rating of the subjects own symptoms than those from plain boxes etc.
    If the doctor assessing the size of the ulcer spoke to the experimental subjects and the subjects perception of their symptoms had been influenced by the type of placebo they’d been given that’d be a possible feedback mechanism for the patients subjective ‘improvement’ to feed back into the supposedly objective medical assessment of ulcer size. Doctors generally want their patients to feel better and believe what their patients tell them. Perhaps an improvement in the patients body language would be enough. So doctors being only human after all, it seems to me there’s a possibility that doc’ll be looking for confirmation of improvement before he even starts putting the pipe in. The experiments could have been designed so the ulcer size ratings were carried out on cctv by a second specialist who never saw or heard from the subjects but blinding the people rating the ulcers size from the experimental subjects who likely have been influenced by the superficial attributes of the placebo they’d been given would be, afaict, an unnecessary complication in trials designed to compare a treatment expected to be effective vs placebo.
    Though of course you’d have to look at the experimental designs (behind the pay wall) for the ulcer trials to know for sure I suppose.
    Hope I’ve made myself clear, let me know if not.

  4. #4 by Stu on April 21, 2010 - 11:25

    I agree with the Doc – good post!

    Richard Dawkins touched on placebo in the God Delusion. I think it was in the chapter/section about people finding comfort in religion (I’ve loaned my copy out so can’t check).

    A reassuring chat with a professional physician can sometimes be enough to make people feel better just by lifting their mood – as the Doc points out.

    This is how woo treatments work – everybody I speak (argue) with about crystal healers etc always, without exception, say “But he/she is a really nice person who understood exactly what was wrong and it worked for me.”

    Doctors – real doctors – have been using placebo to, in my opinion, good effect for a long time.

    It’s still a lie I suppose but a white lie is sometimes justified though isn’t it?

  5. #5 by Mike on April 21, 2010 - 11:43

    To be fair to Moerman, I don’t know that his paper claims that four pills make ulcers heal faster than two pills.

    However, Ben Goldacre promotes Moerman’s findings in his lectures and books, stating (paraphrasing here) “gastric ulcers heal faster when you take four sugar pills instead of two sugar pills, and this is a evidence of the amazing power of the mind over the body”. He is certainly implying causation.

    As I suggest in the article, in light of the NEJM article (plus a Cochrane report which I only became aware of after penning this article) which concludes that placebos have no objective effects, I’m inclined to think that the four pill/two pill effect is actually an artifact of the research process, not a real effect in the patient.

  6. #6 by Jon d on April 22, 2010 - 01:18

    There’s a more recent placebo effect paper by Moerman that you can read for free. http://www.annals.org/content/136/6/471.full if Ben Goldacre had accurately summarised his earlier work I’d sorta expect moerman to say something about 4 placebo pills being better at healing ulcers than 2 but he seems not to. There’s a graph showing that across placebo controlled trials of ranatadine on ulcers, placebo treated ulcers tended to heal better in the trials where the real pills were more effective… But seems to me that could be showing a lot of things apart from placebo effects… Perhaps some trials just simply had more optimistic ulcer measurers than other trials?

(will not be published)