Placebo has strength in numbers

Photo by Flickr user anitacanita. Click for sourceWired has an excellent article on how the placebo effect is increasing in drug trials and how drug companies are trying to understand why. It’s an intriguing article but it conflates two distinct concepts of ‘placebo’ that need to be separated to fully understand the effect.

The term ‘placebo effect’ is used to refer to two things in the medical literature. The first is a statistical concept and it refers to the improvement in patients given an inactive treatment in a drug trial in comparison to those given the actual drug. The second is a psychological concept and it refers to improvement due to expectancy and belief.

If you’re not sure how these are different, you may be surprised to learn that you don’t need a mind to demonstrate the placebo effect – in fact, even rocks can show it.

Let’s say an oil tanker has sunk, the local beach is covered in oil, and you want to compare how effective two cleaning products are – the first, liquid soap, our active treatment, and the second water, our placebo.

So we randomly assign oily stones to a bucket of soapy water or to a bucket of water. It turns out that while stones in the soap condition become less oily, so do stones in the placebo condition, although, perhaps, the effect is weaker. Oil breaks down on its own, water movement disperses it, oxygenation happens. There’s a whole bunch of stuff which means our placebo ‘treats’ the stones.

Statistically we have a placebo effect, because in a trial anything which causes improvement not to do with the active treatment is chalked up to the placebo effect.

In humans, similar effects are at work. Most illnesses improve on their own, when we catch anything at its worst typically it will return to its normal state (an effect known as regression to the mean), people change their behaviour to become more healthy when they’re ill, and so on. None of these are to do with expectancy or beliefs about taking a pill.

But here’s the other thing. Because the statistical concept of placebo is drawn from the study data, the study itself has an effect.

For example, the strength of the placebo effect is measured relative to the active treatment. The Wired article says that placebo is getting stronger, which is another way of saying that the difference between placebo and the drug is getting smaller.

It turns out that the more rigorous the study the less strong the drug effect is, or, in other words, the stronger the placebo effect.

For example, we know that better designed and higher quality studies show smaller drug effects. This includes things as simple as randomisation. If your method for randomly allocating people to groups is more susceptible to bias, it’s more likely to produced biased results. Better randomisation improves the placebo effect, again, nothing to do with expectancy or belief.

So one reason why the placebo effect might be increasing is that studies are just more rigorous these days.

Of course, on top of all of these things, individual psychology plays a part as it adds improvement, and anything which leads to improvement gets captured by the statistical placebo effect.

However, the lab-based studies which investigate placebo look almost exclusively at the psychological placebo effect. They examine the effects of beliefs and expectations but usually carefully control the presence of the unpleasant thing, like pain, so it doesn’t naturally improve and you can’t change your behaviour like you would in real life.

You can’t explain the statistical placebo effect just with psychology. It’s part of it, but not the whole story.

So when I read the article which said that drug companies are busily doing lab studies to understand why the placebo effect is increasing I became a bit suspicious.

The first thing you’d do is look at how your studies have been run, not look at the psychology of belief. Drug companies undoubtedly know this. They’re masters of drug trial sleight-of-hand and know research methods inside out.

The article touches on a likely explanation – marketing. They would like to influence your beliefs so the drug works better for you, because once it’s on the market, it’s the customers’ experience that brings them back for more.

In an industry where genuinely new drugs are rare and most are just no-better copies of rival medications, your beliefs could make all the difference.

Link to Wired on the increasing placebo effect.
Link to previous Mind Hacks post on the psychology of placebo.

11 thoughts on “Placebo has strength in numbers”

  1. Thanks for such a thoughtful post, Vaughn. The reason that “psychological” and statistical placebo effects are conflated in my article is because they’re conflated in the clinical trials — i.e., in the data from placebo control groups. Drug trials, as you probably know, only distinguish between the two types of placebo-related data when they feature a “no treatment” arm, a placebo arm, and an active drug arm — and the vast majority of clinical trials don’t. But I could have made that point more clearly in the article.
    I’m a little hesitant to refer to any kind of placebo effect as “psychological,” because that suggests that it’s related to neurosis or something, just as early placebo researchers like Henry Beecher erroneously believed. It’s almost better to call it a “cognitive” phenomenon, though there are also placebo effects, as I describe in the Benedetti section, that are Pavlovian — created by conditioning — rather than cognitive.
    You wrote: “So when I read the article which said that drug companies are busily doing lab studies to understand why the placebo effect is increasing I became a bit suspicious.”
    Please note that in the article I was *not* saying that drug companies are “busily doing lab studies” to understand what you call the “psychological” placebo effect. I make the point that those studies were primarily done by academics outside of the industry, including the people I quoted in my article, like Fabrizio Benedetti and Tor Wager. In fact, until fairly recently, the drug companies pretty much ignored this research — and at their peril and expense, it turned out. Only now are those two streams of research coming together, specifically because the drug companies are, just as you say, “masters of drug trial sleight-of-hand” — but have still not been able to solve this problem. The under-the-radar collaboration between drug companies that I discuss at the end of the article, the Placebo Response Drug Trials Survey, is just now getting underway.
    Thank you for such an insightful read of the piece!

  2. I don’t quite understand the argument that more rigorous studies decrease the placebo effect. They decrease the active drug effect (which is the active arm effect minus the placebo arm effect) but this is not the same thing as the placebo effect which is variously defined but is something like the placebo arm effect minus a no-treatment arm effect, or even the net change in the placebo arm over time.
    I may be reading you wrong but you seem to have just equated the placebo effect with the inverse of the active drug effect.
    It seems to me that the argument being made in the article is that the placebo arm effect size has been increasing over time – and this is, of course, unaffected by active treatment effect size.
    Now active treatment effect size could be reduced by an increasing placebo arm effect size if they are non-additive, but that’s a slightly different argument to the one you’re making.

  3. Hi PJ,
    You’re quite right and I’m arguing that statistical placebo is relative to the other conditions because the active drug effect, by definition, is its improvement minus placebo (of course, belief and expectancy also contribute to the effect of the active condition as well, so by this definition there is a placebo effect in both arms – another reason to distinguish the definitions).
    This is usually the case because a majority of trials don’t have a ‘no treatment’ arm. As Steve also mentions above, in a three-arm trial with a no treatment condition, placebo can be measured with regards to this. Again, as it’s a statistical result it depends on the trial design.

  4. Nice article. It’s very important to distinguish the methodological issue of non-treatment effects of various kinds from the specific research issues related to response expectancy. There is an independent scientific literature on the effects of expectancy on behavioral responses that does not depend fundamentally on “placebo effect” claims in research methodology. The so-called “placebo effect” as commonly used is a suitcase term for a number of things that include statistical artifacts, spontaneous remission, loose controls, and of course response expectancy as well. As Vaughan pointed out, response expectancy is a factor in both the non-treatment effect size and the treatment effect size, since we expect a result when we get the real treatment as well as when we get the faux treatment.

  5. You’re right that it muddies the water when the same term refers to two different concepts — and that differences in study design can affect the magnitude of the difference between treatments. But I don’t follow the logic of your argument that more rigorous study design necessarily results in a smaller difference between new drugs and placebos. That argument seems to assume that (1) the real drug effect is constant over time and (2) that the bias introduced by poorly designed studies always amplifies differences between drugs and placebos. On the face of it, neither assumption seems warranted. If there’s a problem with randomization, for instance, it could go either way. And isn’t it possible that some of the apparent increase in the placebo effect can be attributed to more trials of drugs that have smaller effects?

  6. “I don’t follow the logic of your argument that more rigorous study design necessarily results in a smaller difference between new drugs and placebos”
    This isn’t my logic, it’s the results of empirical studies (see links).
    While a great susceptibility to bias could lead the study in either way, it typically doesn’t. It typically leads the results to favour the drug under investigation, presumably, because the greater susceptibility means the study is swayed towards the more favourable results for the investigators (typically a drug company).

  7. Great follow up to a very interesting article (though some of definitely goes over my head lol).
    Dumb question, if some people suggest the “placebo effect” works because of the “anticipation” factor of someone anticipating the treatment to work and relieve their ailment would knowing that the placebo effect works be enough to “trick” your mind? Meaning if I knew the placebo effect can work as a treatment in itself, could I knowingly take a sugar pill to treat my back ache and in knowing the placebo effect can work itself cause the placebo effect to work?
    Wow, that went longer than I intended. It kind of makes sense in my head and now just looks really gobbled in written form lol.
    Is it just a dumb idea/question?
    Thanks for the great article and post Steve and Vaughn

    1. No. It wouldn’t work as well because you know you are not giving yourfelf the real pill but if someone else was to trick you it would work.
      What if they’re doing it right now?????

  8. There’s no reason why regression to the mean would differentially affect one group more than the other. Therefore, I don’t see how this is an explanation for the placebo effect.

    When you write about the “active” and “inactive” treatments you imply from the beginning that placebos are inactive. For example, water is not an “inactive” treatment in your example. Rather, you are simply comparing two different types of treatments – one of which happens to be more effective than the other, but fundamentally both are somewhat effective.

    Beliefs and expectations modify the brain activity, which can change the signals to the immune system mediated by the autonomic nervous system and neuroendocrine system. Therefore, it’s not just about “psychology” but a true relationship between thoughts and emotions, brain activity, neural and endocrine activity and ultimately immune function.

  9. What worries one on the placebo effect, is could it be used to generate illness in people rather then cure illness. Stands to reason that if it can be used to cure, it can be used to create. And those with the power to create illness can then market false drugs to cure a non existant illness.

    I have noticed in Africa there is a high HIV infection. While there is no denying that man has found such a virus, could it be that many who die from Aids and are never tested for the virus, be dying because of some placebo effect. A classic case of the body can not live without the mind.

    If you tell a mind it is dying from some dreaded disease which one does not have, could the mind then go on to die killing the body with it. Given HIV is a sexual disease, it does bring to light other facts, like how can United Kingdom have the highest teenage pregnancy cases in the world(proof protection is not being used in the young) but yet remain so low in HIV contraction.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s