350 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS effects and safety under conditions of use must be as close to 100% as possible. As a result, the manufacturer of cosmetics needs reliable information on the probability of adverse reactions to his product in advance o[ its marketing. Cosmetics in general are a fairly innocuous class of products' if they were not, the manufacturers would have had to close their doors long ago. The fact that billions of dollars are spent annually on cos- metics confirms that consumers as a whole do not find cosmetics irritating. The skin of man has been constantly exposed to a wide variety of dif- ferent chemicals. The milieu of our modern industrial society further insults our skin with a wide range of chemicals in the form of clothing, industrial finishes, and dusts. Experience has made the elimination of known skin offenders from cosmetics a fairly simple task. Most potential irritants have been identified, and their use in cosmetics is almost nonex- istent. Kligman (1) pointed out, on the other hand, that "practically all substances are capable of being contact sensitizers for some persons under some conditions." However, the exclusion of all chemicals which--under one condition or other-could act as irritants or sensitizers would make the formulation of cosmetics impossible. With this in mind it becomes apparent why an evaluation of currently available predictive test procedures is important. A description of pre- dictive testing techniques, especially as they are applied to cosmetics, has recently been made by Brunner (2). A thoroughly annotated review of this subject has been prepared by Idson (3). In addition, a well-reasoned critique of standard test methods for cutaneous contact allergy has been published by Kligman (1), who concludes that the techniques are "insen- sitive" to moderately strong known sensitizers. In view of this he de- veloped his so called "maximization test" (4). In spite of some obvious deficiencies, predictive testing techniques for cosmetics appear to be reasonably reliable. If they were not, the inci- dence of reaction to newly introduced cosmetics would be expected to be much more common in our industry. In fact, we feel that a well-executed predictive program is most useful in keeping potentially troublesome products off the market. The purpose of this discussion is not to suggest new tests. Instead, the need for careful selection of predictive testing procedures and for judicious interpretation of the resulting data will be pointed out. In ad- dition, some problems which may be encountered during predictive test- ing will be explored.

PROBLEMS OF PREDICTIVE TESTING 351 PROBLEMS OF PREDICTIVE TEST PROCEDURES Availability of Test Procedures The existing techniques are widely used, and all ot5 them can help in predicting the safety of cosmetics if they are wisely selected and if the data obtained are properly applied. These procedures can and should be modified or exaggerated to suit the product however, the application of results from a small-scale test to the population at large is statistically dif- ficult (5, 6) and leaves much to be desired, especially in view of the artifi- cial conditions which are employed in testing cosmetics. It is probably worthwhile to mention briefly the various groups of tests that are available. Animal tests may include the guinea pig immersion procedure (7), the "Draize" rabbit eye test (8), the standard "Draize" dermal irritation pro- cedure (8, 9), various guinea pig sensitization tests (8, 10, 11), and others. Human patch tests in a variety of forms have been used for many years. These may include the single (primary irritation) patch test (1, 3), the prophetic patch test (1), modification of the patch test to determine possible phototoxicity or photosensitization (3), repeated insult patch tests to determine sensitization (12-14) and, finally, the maximization test procedure of Kligman (1, 4, 15, 16). In-use tests represent the third group of tests available (17). Although no formal procedures for this type of testing exist, the product is generally used in accordance with the manufacturer's direction after preclinical or clinical patch tests. In some instances, it may be desirable to exaggerate the conditions of use by increasing the amount and frequency of applica- tion. Preferably, the period of intensive use is followed by a rest period and then a "challenging" patch test. Problems of Selection The protocol 15or an effective predictive testing program requires a considerable amount of judgment. One of the most important limita- tions on test procedures is the fact that not all of them are applicable to all products, and the selection of the test is as important as the test itself. To subscribe to the concept that a product must pass a particular test be- [ore it can be considered safe for marketing eliminates all opportunity to modify a standard test. It also precludes the need for interpretation of the results obtained by the clinician and, in effect, voids his expert judg- ment.