456 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS The second question may usually be answered by comparison in a clinical trial but ethical considerations require that new treatments are admitted to a trial only if there are good grounds for expecting them to be as least as effective as the old treatment. The first question is more subject to ethical difficulties in the answering, since only in the mildest diseases may treatment be stopped in order to study the effect of no treatment. The ethical considerations referred to above are not unequivocably interpreted, as has been pointed out by Hill (1), since the only well estab- lished treatment may actually be inferior to the absence of treatment, as in the use of anti-coagulant in cerebrovascular disturbances. No scientist is completely free from the ethical implications of his work and this applies to the statistician as much as to any other scientist. We may, however, consider the scientific aspects of a problem without reference to its ethical context. When attempting to answer the first of the two questions above, a placebo is frequently administered instead of merely neglecting those patients who are to receive no treatment. One of the original meanings of placebo is derived from the same root as placate, that is the placebo is given in order to please the patient or to keep him from the knowledge that he is being neglected. Lasagna (2) has given several meanings for placebo we are at present concerned with the administration of a dummy treatment (3), which may also and incidentally please the patient either consciously or in more subtle ways. There is little to be gained by distinguishing form- ally between placebos and dummies (4). QUANTITATIVE RESPONSES TO DRUGS When a drug or medicament is administered to a patient the response is likely to be quantitative if the treatment has pharmacological properties but it may be quantitative or qualitative if the effect of the treatment is psychological or psychosomatic. The effect of a pharmacological preparation on a living organism must depend on the potency of the preparation and on the dose used but the mode of observation of the effect may either be quantitative as well or it may only be possible to observe success and failure of the treatment. The psychological results of the administration are likely to be qualitative the treatment either pleases or it does not, although degrees of pleasure may sometimes occur. Both pharmacological and psychological effects are probably present in most clinical trials, each to a greater or lesser extent. If the observed

THE EVALUATION OF PLACEBOS IN CLINICAL TRIALS 457 response is quantitative we may suppose that pharmacological factors predominate although this is not always the case since, for example, Wolf et al (5) found it possible to affect the eosinophile count by placebo action. Such quantitative observations would occur, for example, in a thai to test the effects of treatments by various fruit preparations on the vitamin C content of the blood. These effects can only be measured against a stan- dard which in this case consists of the absence of treatment. However those persons receiving the fruit preparations are thereby also receiving extra water, sugar and other components besides vitamin C so that the absence of treatment should be interpreted as receiving equivalent amounts of water, sugar, etc., and perhaps even sulphur dioxide, rather than as complete neglect. At the same time any psychological factors will be equalized between the active treatments and the blank treatment or placebo. The function of the placebo experiment, in such a thai is to avoid bias in the results just as in, say, the direct determination of oxygen in rubber by the Unterzaucher method (6) it is necessary to do a blank deter- mination, or as in absorption spectrometry the intensities must be corrected for the background absorption. An important additional characteristic of the clinical trial is the pronounced variability between individuals receiving the treatments. It is therefore particularly important to design the thai so as to control and reduce sampling error especially by the random allocation of treatments and by ensuring that the individuals studied form a random and representative sample of the population for which the treatments are intended. A comprehensive account of the application of statistics to the design and analysis of clinical trials has been given by Hill (7) and the general principles of experimental design have been discussed nonmathe- matically by Cox (8). QUANTAL RESPONSES TO DRUGS If the observed response of an individual to drugs or other medicaments is success or failure of the treatment the response is termed quantal. The object of the treatment of serious diseases is normally the survival of the patient, and chronic conditions usually require the alleviation of pain or other symptoms. In the first case success denotes survival, and in the second improvement, versus death and no improvement respectively. The reaction of a patient to the administration of pharmacologically inert substances is usually referred to as placebo reaction. This reaction is normally quantal so that in clinical trials where there is a possibility of