WHAT ABOUT DEPRESSION?
Although the 1970 study on tablet colour suggested that depression, too, was placebo-responsive, hard data is thin on the ground. Hundreds of clinical trials have compared the effects of antidepressant medication with those of placebos, but hardly any have included a no-placebo group. In the absence of such studies, we cannot be sure whether or not depression is placebo-responsive. In 1998, however, two American psychologists, Irving Kirsch and Guy Sapirstein, came up with an ingenious way round this problem. None of the drug trials they looked at included a no-treatment control group, but a whole batch of studies of psychotherapeutic treatments did. Many of these studies, for example, compared the recovery rates of depressed patients receiving psychotherapy with the recovery rates of other depressed patients who remained on waiting lists. Kirsch and Sapirstein used the waiting-list recovery rate as a rough measure of the spontaneous remission rate, and compared it with the recovery rate shown by those taking placebos in the drug trials. In this way, they were able to estimate the relative effects of placebos and antidepressant drugs. Their conclusions were startling. Those taking drugs showed, on average, about 33 per cent more improvement than those treated with a placebo. But those taking a placebo showed around 200 per cent more improvement than those who received no treatment at all. If we assume that the placebo component of the real drug is the same as that of the pure placebo, then we must conclude that around 25 per cent of the improvement shown by those taking antidepressants is due to spontaneous remission, 50 per cent to the placebo effect, and only a measly 25 per cent to the antidepressant medication itself.17
Donald Klein, a psychiatrist at Columbia University in New York, has cast doubt on this dramatic conclusion.18 Klein argues that many of the studies on which Kirsch and Sapirstein based their claims were flawed for various reasons. More importantly, Klein also raises questions about the scales used to measure recovery. A typical five-point scale might go something like this:
1 = normal functioning
2 = mild functional impairment
3 = definitely impaired
4 = cannot work
5 = impaired self-care
Now, suppose that all the patients in a trial start out at a mean of 4.5 on this scale – they cannot work, and some of them cannot even take care of themselves properly. If the average patient on medication improves by three points, and the average patient on placebo improves by 2.25 points, Kirsch and Sapirstein would argue that for the typical patient, 75 per cent of the drug effect is attributable to the placebo effect, which makes the superiority of the drug seem relatively trivial. If, however, we look at where the patients end up on the scale, the superiority of the drug begins to look more important, for the average patient on medication is now between normal functioning and mild impairment, while the average patient on placebo is now between mild and definite impairment. This difference is clinically important, and yet it is hidden by the way that Kirsch and Sapirstein report the data.
On the other hand, it may be that the placebo effect in depression is even greater than Kirsch and Sapirstein claim. As they themselves point out, the placebo component of a real drug may be higher than the placebo effect of a pure placebo. Placebo effects depend in large part on the belief that you are being given an effective treatment, and there is evidence that many participants in clinical trials guess correctly whether they are receiving the real drug or the placebo. The real drug tends to have noticeable side-effects that the placebo lacks. In one study comparing two antidepressants (imipramine and phenelzine) with a placebo, 78 per cent of patients and 87 per cent of psychiatrists correctly guessed who was receiving an active drug and who was receiving the placebo.19 In fact, in twenty-three out of twenty-six studies where researchers bothered to check, both patients and physicians did better than chance at guessing who was receiving the placebo and who was not.20
If patients guess they are taking a placebo, this may reduce the power of the placebo to unleash their internal healing resources. On the other hand, if they guess they are taking the experimental drug, this might enhance the placebo effect. Arguing along these lines, Kirsch and Sapirstein suggested that some of the apparent difference between real drugs and pure placebos may be due to an ‘enhanced placebo effect’. If this is true, the ratio of the specific effect of the real drug to its placebo component becomes even smaller still.
To disentangle the true specific effect of a treatment not only from spontaneous remission and the effect of a pure placebo such as a sugar pill, but also from the enhanced placebo effect of the real drug, Kirsch and Sapirstein suggest that clinical trials should involve not two arms, as most do at present, nor even three, but four. In addition to the experimental group, the no-treatment group and the pure placebo group, a fourth arm would treat patients with a so-called ‘active placebo’. This is a placebo that lacks the key ingredient of the experimental treatment but contains some other active substance designed to produce similar side-effects. Some trials of antidepressants, for example, have used atropine as an active placebo because it produces one of the most noticeable side-effects that many antidepressants do – a dry mouth – but does not have any specific activity against depression itself. Needless to say, such four-arm trials are even less common than three-arm trials. In fact, they are virtually non-existent at present. They would, however, be invaluable in advancing our understanding of the placebo phenomenon.
A few clinical trials of antidepressant drugs have used active placebos instead of pure placebos. That is, they have compared two groups of patients – one that receives the experimental drug, and one that is given an active placebo such as atropine. These trials reduce the chances that the double-blind will be broken – in other words, they make it harder for patients or their doctors to guess who is receiving the experimental drug and who is receiving the placebo. Interestingly, these trials tend to show much lower drug effects than those in which antidepressants are compared to pure placebos.21 This, in itself, provides some evidence for a placebo effect in depression. The argument is a subtle one, but worth exploring.
If depression was not placebo-responsive, then the whole effect of antidepressant drugs would be due to the pharmacological properties of the drugs themselves. It would make no difference whether you believed you were taking the drug or not. The difference in recovery rates between patients taking the drug and patients taking placebos would therefore be the same whether or not patients were able to guess correctly which group they were in. This is not, however, what we find. When patients are able to guess which group they are in, the difference between the experimental group and the placebo group increases. This suggests that belief is indeed playing some role in the efficacy of antidepressant medication. The placebo effect, in other words, does seem to work for depression.
PLACEBOS AND SCHIZOPHRENIA
If placebos worked for every kind of mental disorder, their power would not be limited to anxiety and depression, but would extend even to the most severe psychoses, including schizophrenia. Is there any evidence that placebos can affect such extreme mental conditions?
In the 1950s, many psychiatrists were unwilling to accept the idea that placebos could alleviate depression, let alone schizophrenia. One German psychiatrist, however, decided to test the idea. Heinz Lehmann picked three of the most mute and deteriorated schizophrenics on one of the back wards at Verdun hospital. He told the patients and their nurses that they were going