PREDICTIVE PATCH TESTING 329 "skin fatigue" is not a factor. However, one may question the advis- ability of using occlusion or other methods of facilitating penetration, such as skin damage with sodium lauryl sulfate or freezing, in predictive testing. One of the reasons that a formulation does not sensitize in ac- tual use may be that it does not penetrate, and a false impression of allergenicity may result from artificially making the substance pass through the skin in the test procedure. Thus, alterations in any of the test conditions can change the inci- dence of sensitizations. Of major interest, however, is not the rather academic question of what the potential sensitizing power of a substance or formulation may be when exploited to its limits by the most stringent conditions but how it will behave under use conditions. Can one fore- cast, for example, what the incidence of sensitization will be in consumer use of a hair dye (used every six weeks on the scalp, without occlusion, with exposure to the dye intermediates for a minute or two) when 1 in 50 test subjects reaci:s after 10 or 15 daily applications to the arm under oc- clusion? There is, as yet, no predictable relationship between the two situations. If there is a high yield of true reactors (for example, more than 20%) on any of the predictive tests, one may conclude that the sub- stance is probably going to be troublesome regardless of the conditions of contact, but especially if severe auxiliary damaging stimuli have not been used. Low sensitization rates, in the range of 1-5%, are generally ac- ceptable, particularly if this incidence is no greater than that of a known control which has already had extensive consumer use, with a tolerable level of sensitization reactions. The control must be used in the same way as the new agent and have similar features as regards penetrability, irritancy, etc. The level of sensitivity of the test should probably be fixed at the point at which the control agent (which is safe for consumer use) produces at least an occasional reaction in the test. Otherwise, the sensitivity of the test may be too low to detect potentially troublesome materials. For different classes of cosmetics it may be necessary to mod- ify test conditions, such as frequency and duration of application, occlu- sion, etc. The Brunner-Smiljavic test, for example, was designed for use with mercaptan-containing hair waving formulations which are applied unoccluded for 30-60 minutes in consumer use. Alteration in time and type of contact is needed for some other types of products, such as facial creams. With new classes of compounds animal tests may be useful for prelim- inary screening before any human sensitization testing. The procedure described by Voss (9) appears satisfactory for this purpose. Based on

330 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS his results, one may conclude that, if a substance produces a significant number of reactions in the guinea pig, it is also likdy to produce sensiti- zation in human testing. However, some agents produced no reactions in the guinea pig but did sensitize humans. A rational program may, therefore, involve consecutively the prdiminary animal screening test, followed by an applicable human sensitization test, and finally a super- vised use test by a panel of subjects. With consumer usage panels, dependable confirmatory answers on sensitization may be expected only if the panel is large, the contacts frequent, and the observation close and prolonged. Valuable information can also be obtained from market tests, which are the final step before unrestricted sales, but there must be careful follow-up of users, and the trial sale period must be long enough to allow for the repetition of contacts which is a requisite for sensitization. A trial sale period which is only long enough to allow for one or two uses of a product is inadequate for a study of sensitization. The general principles of predictive testing for sensitization can be recapitulated as follows: To obtain a positive control an agent is applied to the skin in such a manner that it will produce allergic sensitization reactions in one or more subjects in a small test group, whereas in actual use it produces only a negligible number of reactions. If other agents applied in the same way give no more reactions than this control, it is assumed that they will be no worse in actual use than the control they are, therefore, acceptable as regards sensitizing potential. It is the dif- ference between test conditions and usage conditions which changes the sensitization rate. Aside from occlusion, which has an obvious effect on penetration, the main differences are amount of antigen applied, duration o[ contact, frequency of contact, spacing between contacts, and total ddration of exposure. It is generally held that on contact the antigen combines with some component of skin protein, is absorbed and reaches the reticulo-endothelial system, where it causes certain cells to produce antibodies. The lymph nodes draining the area of skin to which the sub- stance is applied are the important reticulo-endothelial tissues in this type of sensitization. In experimental sensitization, Cainan and co-workers (5) showed that varying the site of application reduces sensitizing poten- tial, since this apparently reduces transport of antigen to a single group of lymph nodes to a level below the threshold required to cause modifica- tion of lymphold cells and the formation of antibodies. Why lengthen- ing of the interval between applications appears to decrease the per- centage of reactors is not clear. Perhaps, closely spaced repetition results in an accumulation of antigen in regional lymph nodes to that