VEHICLE EFFECT ON ANTI-PERSPIRANT ACTIVITY 187 Table VIII Anova of Post-Test Data Source of Variations DF SS MS F Value* Treatments (A) 3 1.411 0.470 5.311 Subject Pairs (B) 3 0.659 0.220 2.486 Weeks (C) 3 0.079 0.026 0.294 Sides (D) 1 0.004 0.004 0.125 Subjects x Sides (BD) 3 1.002 0.334 10.438 Weeks x Sides (CD) 3 0.616 0.205 6.406 Treatments x Sides (AD) 3 0.145 0.048 1.500 Pooled Residual 12 0.721 0.060 SS^B. 6 0.531 0.0885 (error (A)) SSAm), 6 .190 0.0320 (error (B)) Total 31 4.637 *where F3,6, o5 = 4.76 F•,6,.o 5 = 5.99 The pretest blocking procedure was instituted to account for the light- or heavy- sweating side treated, yet the sides' (d) extremely low F value indicated that the side which was treated could have been selected randomly without regard to pretest sweat ratios. This is in agreement with Wooding and Finklestein's (14) decision to eliminate pretest sweat ratios in evaluating antiperspirant efficacy. This is further substantiated by the insignificant F value of subject pairs (B). Treatment by sides effects (AD) indicates that the side to which the treatment is applied has little effect on the treatment's efficacy. However, the F value indicates a strong subjects by sides (BD) interaction. Figure 1 gives an indication of how this interaction occurred. All subject pairs exhibited similar sides by subjects effects, where the average log ratio is greater when the light-sweating side is treated, except for subject pair 3. The daily pretest sweat ratios for this pair showed extreme day-to-day variation, but averaged close to unity, so designation of one side as heavy or light-sweating for members of this pair was subject to error. Otherwise, the other three subject pairs had no difference. As expected, there was no significant difference in the effects of weeks (C), despite the fact that during the first three treatment weeks, the average environmental temperature was in the low 60'søF and the average relative humidity was 40%, while in treatment week 4, the average temperature was in the high 70'søF and the average relative humidity was 70% (18). Figure 2 shows the major change during week 3 in heavy versus light-side treated. This does not appear to have been caused by environmental factors, but may have resulted from emotional factors. Treatment week 3 occurred during a period when there was a great deal of pressure on the subjects, who were all in academic surroundings, to finish course work and write exams. Emotional sweating has been suggested as a cause of erratic sweat ratios and rates (19). However, it did not affect the antiperspirants' effectiveness. All subjects participated in muscular exercise, such as tennis, jogging, or dancing, at various times. This activity was noticeable in daily sweat collection weights, but not in overall sweat ratios. Antiperspirant effectiveness was maintained during exercise. Exercise and emotional stress had the greatest effect on sweat weights.

188 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS (( (eOkp ,) kp 1.0 L O. 14 # L L/H O.S 0.6 0.7 0.! 0.9 1.0 B • 4 I Figure 1. The pretest sweat ratios of light-sweating side to heavy-sweating side (L/H) of Subject Pairs (B) versus Subjects by Sides (23.•,(BDkp) ) illustrates the Subjects by Sides interaction occurs in Subject Pair 3. H = Heavy-sweating side treated. L = Light-sweating side treated. A significant difference in treatments was caused by the stick antiperspirant. The LSD test verified this and also showed there were no significant differences among treatments 2, 3, and 4. Percent sweat reduction results (Table VI and VII) in decreasing order of antiperspirant effectiveness response are treatments 3 (lotion, 38.0%), 4 (hydroalcoholic solution, 32.4%), 2 (cream, 31.8%), and 1 (stick, -6.2%). The results of the treatments containing Chlorhydrol ©, 50% w/w solution, are within the FDA-OTC Antiperspirant Panel's effectiveness percentage range for these vehicles. Except for the stick formulation, they also exceed the panel's guideline that a product must reduce perspiration by 20% to qualify as an antiperspirant (20). The Latin Square design was statistically satisfactory in evaluating antiperspirant efficacy because it provided statistically useful results with a small number of subjects, permitted the testing of each subject with each treatment over a relatively short period of time, and eliminated subject pair or weeks effects in comparison of treatments. Since the results indicated that side treated may be selected randomly without regard to heavy or light-sweating sides, a regular Latin Square instead of a split plot could be used in future testing.