THE CLINICAL EVALUATION OF ANTIDANDRUFF SHAMPOOS 441 To overcome this difficulty, it is essential to compare the progress of a treated panel with an untreated panel running concurrently. Dandruff investigations would be simplified considerably if objective measurements were practical. A technique has, in fact, been published whereby scale samples are taken by means of a miniature vacuum cleaner and subsequently weighed (4). Since this technique may involve the dis- turbance of scale attached to the scalp which may influence the disorder itself and in the absence of alternatives not showing a similar disadvantage, a subjective method of evaluation is to be preferred. Clinical trials in which subjective assessments are made always embody the risk of significant observer error and bias. Comparison between treat- ment and no-treatment panels tends to rule out error and bias due to the subjects themselves, especially if they are allocated to the different panels by a suitable method of randomization but it is still vital that the observer should not be able to identify whether a subject is receiving treatment or not at the time of inspection. For various reasons it is preferable that this knowledge should also not be available to the actual subject and so the "double blind" technique should be adopted. This involves the employment of a placebo, identical in all discernible respects to the treatment but lacking the active constituents one half of the subjects receive the treatment and the remainder have the placebo and serve as controls. Whilst it is usually possible to ensure a close resemblance between a placebo and treatment ("control" and "test") shampoo, some ingenuity of the formulator is sometimes needed to achieve this. To make sure that the identity of test and control materials is unknown to both subjects and observers, whilst also ensuring that the observers do not know or try to guess the allocation of subjects to the various panels, it is desirable to apply several different code-letters both to the test and control products and to allocate them to the subjects by means of a randomi- zation chart. This also helps to prevent subjects comparing notes with one another and possibly influencing their cooperation. Elimination of bias may also be assisted by stratifying the subjects according to various criteria possibly affecting the dandruff condition, e.g. ensuring •hat various degrees of initial severity are equally distributed between the panels. Age and sex might well be treated similarly, though the factor next in importance to initial severity is probably the question of prior usage of a medicated shampoo. "Cross-over" technique would theoretically help greatly to strengthen the validity of a clinical trial that is, one panel would start with the control

442 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS product and later use the test, whilst the other panel would use the two products in reverse order. If, however, the test product has any real effect, it is vital to know how soon the dandruff level reverts to normal when treatment is discontinued, otherwise a "carry-over" effect will be operative during the control phase for the panel starting with the test product, results during the control phase will also depend on the actual time required to regenerate dandruff by infection or other means. The post-treatment zone illustrated in Fig. 2 shows that several weeks are required to re- establish the original pre-treatment dandruff level when an active anti- dandruff agent has been used. A practical solution might be to allow a prolonged time-lag between test and control phases, but this is likely to prolong the whole procedure of experiment to an unacceptable extent. A "cross-over" trial ignoring the points noted would tend to underestimate the efficacy of a treatment. o i i , 5 10 I PIE-TREATMENT IN WEEKS I I TREATMENT I I_.. {TESI Figure 7 Results of 16 weeks' treatment with shampoo containing 2 % w/w zinc bis(pyridine-2-thiol l-oxide) on a case of moderate-severe dandruff. (Detailed method of inspection). It is almost a truism to suggest that the statistician who will have to determine the significance of the results, should be fully consulted at the planning stage of a clinical trial. It is the authors' view that a carefully- planned and well-conducted trial on a limited panel is far more valuable than a poorly-conceived investigation carried out on a relatively vast scale an essential feature of sound planning is to ensure that the results will be adequate and in a suitable form for statistical analysis.