356 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS are unequivocal: The product causes adverse effects or it does not. The in-use test still has many important limitations, such as: number of sub- jects their age and sex period of use geographical locale and possible lack of sensitivity to misuse or deliberate abuse. These problems can sometimes be resolved by careful consideration of the product's intended use and the manufacturer's recommended directions. A major problem is the reliability of the test subjects who must be able to communicate with the investigator and must be depended on to use the product. This problem has been noted by Maibach and Epstein (12), who feel that such a test is frequently not a test at all. This is, of course, a matter of experi- mental design, but does not detract from the value of properly supervised and controlled in-use tests. CLINICAL EXPERIENCES So far, the discussion has been primarily in terms of the possible chance of error or possible problems in the interpretation of results. In order to illustrate these problems and to re-emphasize the need for care- fully controlled in-use tests, two examples from the authors' experiences will be discussed. Approximately six years ago a patch test study of a make-up product elicited no evidence of either primary irritation or of sensitization via the repeated insult technique. More recently, this composition was re- examined for primary irritation and sensitivity by the same technique at a different test locality. Surprisingly, the product now caused primary ir- ritation and could not even be subjected to the repeated insult test be- cause of the high incidence of irritation. Our approach to this riddle was a pragmatic one: A series of break-down products, in which one or more ingredients were deleted, was subjected to patch and sensitization tests. After much effort, the offending ingredient was identified, and the product was then reformulated to yield a new, hopefully safer composi- tion. Although a definitive explanation cannot be offered, several things could have taken place during the intervening five or six years: There may have been subtle changes in the raw material due to a different process of manufacture. The time of year during which the tests were conducted was different. The selection of test subjects could have played a role in causing the second test to show positive results, whereas six years ago the product was considered safe and ready for marketing. Last,

PROBLEMS OF PREDICTIVE TESTING 357 but not least, the change in testing facilities may have been the influenc- ing factor. Interestingly enough, the incidence of complaints from con- sumers did not change during the intervening years. What does this all mean? Are our tests too sophisticated or are the artificial test conditions too severe and thus prejudicial to the product? Of course, we do not know the answer, although Baer's (23) report on the changing sensitivity of the population may offer a clue. Regardless, this is a typical example of how the responses to chance patch testing can trig- ger research activities.. A more interesting example is a fairly recent experience using a trans- lucent facial make-up product. This product was put through the bat- tery of tests using established patch test methods and found to be free of irritation and sensitizing liability. However, in an in-use test, over 75% of the test panelists were unable to tolerate the test product. At first it was believed that the geographical location of the test site might have pre- cipitated the responses. Alternately, exposure to natural sunlight could have caused the reactions. Accordingly, photopatch tests (on the back) were conducted with negative results. Another in-use test was initiated at another site, which confirmed the results of the first in-use test. In view of these adverse results, the product was reformulated in an attempt to eliminate the offending agent or agents. After much work, some minor constituents were eliminated from the formulation, and a third in-use test was conducted. Now, the revised product was tolerated by 100% of the test population. It was concluded that neither light nor oc- clusion triggered the reaction. In this case it seems evident that the lo- cation of the skin, i.e., face vs. back, was the primary factor in producing the adverse effects. If the initial patch test results had been used as the sole criteria in judging the product safe, the original product would have been marketed on the basis of established procedures and in good faith. , The product would, nevertheless, have caused adverse reactions in many consumers. Based on this example, exaggerated in-use testing would seem to be an important part of any predictive skin testing program. It should be emphasized that the two examples cited here are unusual. Normally, repetitive patch tests of the same product yield comparable results. Similarly, in-use tests normally confirm the safeness of products established in a patch test series. Our many years of experience indicate that a well-designed and well-conducted predictive testing program is an almost foolproof method of establishing the safety of products before they are marketed.