Placebo: Mind over Matter in Modern Medicine. Dylan Evans
and requoted by hundreds of doctors who never read Beecher’s paper, let alone the studies which he reported. The experimental evidence was trumped by the word of authority, in a parody of the whole enterprise of scientific research. The founders of the Royal Society would have turned in their graves.
Chapter 2 WHAT CAN PLACEBOS REALLY DO?
Dramatic claims have been made for placebos. According to Dr Robert Buckman and Karl Sabbagh, ‘they seem to have some effect on almost every symptom known to mankind’.1 At the other extreme there are those who argue that the placebo effect is largely or even totally illusory. Arthur Shapiro, who spent forty years researching the topic from the mid-1950s until his death in 1995, concluded that there was little evidence for the view that placebos could have a direct and permanent effect on medical disorders.2 Gunver Kienle and Helmut Kiene have probed the literature on placebos in great depth and found it to be full of misquotation, blind repetition of poorly substantiated claims and the uncritical reporting of anecdotes.3 The placebo effect, they claim, is no more than a myth.
So much for the claims; what of the evidence? It is true that placebos have been used in thousands of clinical trials, but – as we saw in the last chapter – most of these studies do not include a no-treatment group. As a result, we cannot be sure that the placebo made any difference. The improvement shown by the patients in the placebo group might have occurred anyway as they recovered their health naturally, even if they hadn’t received a dummy treatment. To discover what medical conditions the placebo response can really affect, we need to look at the research much more carefully. Only if it can be shown that people with a particular condition do better when treated with a placebo than when not treated at all can we be sure that the placebo response really works for that condition.
THE POWERLESS PLACEBO
In the late 1990s, two medical researchers at the University of Copenhagen attempted to settle the debate about the placebo effect once and for all. Asbjorn Hrobjartsson and Peter Gotzsche combed through the medical literature much more extensively than anyone had done before, picking out all the studies they could find that included both a placebo group and a no-treatment group.4 They were able to identify a surprisingly large number of such trials – 130 in all. Of these, 114 provided relevant data enabling a proper comparison of the placebo group with the no-treatment group. Using meta-analysis, Hrobjartsson and Gotzsche pooled the results of these studies and concluded that, overall, there was little evidence that placebos had any powerful clinical effects.
This simple conclusion was seized on by the media and reported as proof that the placebo effect was a myth. If they had read the whole study, however, they might not have been so quick to buy Hrobjartsson and Gotzsche’s take-home message. The devil, as always, was in the details. For one thing, the studies examined by Hrobjartsson and Gotzsche fell into two distinct groups. Some had reported their results in binary terms (such as positive versus negative result), while others had used a continuous scale (such as the amount of pain relief). For the binary group, there was a small placebo effect, but the result was not significant by the normal standards of statistical research. So far, then, Hrobjartsson and Gotzsche were justified in saying that there was little evidence that placebos had any effect. For the studies using continuous measures, however, there was a significant beneficial placebo effect. These studies, then, do provide good evidence that placebos can produce clinical benefits.
Furthermore, the range of medical problems covered by the 114 studies analysed by Hrobjartsson and Gotzsche was enormous. In total, forty clinical conditions were examined, from asthma and smoking to menopause, marital discord and schizophrenia. Hrobjartsson and Gotzsche averaged over all these studies and, because there were relatively few in this sample that provided evidence in favour of the placebo effect, the negative view prevailed. But if you did the same thing for virtually any powerful drug, the result would be the same. This is because any kind of therapy that works – be it a drug, a surgical intervention or behavioural therapy – will help people with some conditions and not others. There is no such thing as a universal remedy, a real-life cure-all, a panacea.
Certainly, some people have claimed that placebos are just this. Beecher was largely responsible for floating the idea that placebos can affect virtually every medical condition. Although the evidence on which he based this claim was – as we have seen – deeply flawed, the myth of the all-powerful placebo soon became the established medical wisdom. If Hrobjartsson and Gotzsche had contented themselves with exposing this myth, the path would have been opened for a more realistic assessment of the placebo effect, distinguishing between those conditions that are placebo-responsive and those that are not. But Hrobjartsson and Gotzsche went further, asserting that there was no evidence that placebos had any effects at all.
This, at least was the upshot of their brief conclusion. In the small print, however, they were forced to concede that in some cases there were noticeable placebo effects. For some conditions such as anxiety the results were too variable to allow a simple interpretation. For all sorts of pain, however, there was clear positive evidence of a significant placebo effect. Headaches, postoperative pain and sore knees could all be relieved by a sugar pill. There was, then, some reason to suspect that, in pooling the results of studies involving so many different kinds of medical condition, the true profile of the placebo response was obscured.
The mismatch between the complexity of the data analysed by Hrobjartsson and Gotzsche and the stark simplicity of their conclusion is yet another reminder of the need for caution in getting to grips with the research on placebos. What promised to be the final, definitive word on placebos turned out to be a poor study, full of flaws and capped by an inaccurate summary. If we are to get a good idea of what placebos work for – if, indeed, they work for anything – it seems we must go back to square one again, and look at the evidence bit by bit. Only in this way, proceeding carefully, can we begin to build up a picture of what placebos can really do.
NATURAL BORN PAINKILLERS
‘To talk about placebos,’ writes the American gastroenterologist Howard Spiro, ‘is to talk in large part about pain.’5 Of all the claims made for the placebo response, those that emphasise its power to relieve pain are the most well-established. The pioneers of placebo research focused almost exclusively on the painkilling properties of placebos. Most of the studies conducted by Beecher, Lasagna and others in the late 1940s and early 1950s were marred, however, by their failure to include no-treatment groups. Typically, they would report that a certain number of patients experienced pain relief after being given a placebo, and conclude that this had been caused by the placebo. This conclusion cannot be trusted, since no attempt was made to measure the spontaneous improvement in patients who did not receive a placebo.
Fortunately, more recent studies of placebo analgesia have included no-treatment control groups in addition to the experimental group and the placebo group. In most of these studies, the no-treatment group has been found to do significantly worse than both the experimental and the placebo groups. We can be confident, then, that the pain reduction experienced by those given the placebo would not simply have happened anyway. Placebo analgesia is real.
In a particularly striking study, patients who had undergone tooth extraction were treated with ultrasound to reduce the postoperative pain.6 Unknown to both doctors and patients, the experimenters had fiddled with the machine, and half the patients never received the ultrasound. Since ultrasound consists of sound waves of very high frequency – so high, in fact, that they are inaudible to the human ear – there was no way for the doctors or the patients to tell whether or not the machine was emitting the sound waves; the test was truly double-blind. After their jaws were massaged with the ultrasound