j. Soc. Cosmet. Chem., 44, 13-21 (January/February 1993) A comparison of antiperspirant data analysis methods FRANK H. DIETRICH II, JAMES P. BOWMAN, BARBARA M. FATH, and JOHN E. WILD, Hill Top Research, Inc., Cincinnati, OH 45147 (F.H.D., J.P.B., B.M.F., J.E.W.), and Northern Kentucky University, Highland Heights, KY 41076 (F.H.D.). Received January 22, 1992. Synopsis The purpose of an antiperspirant study may be to estimate the efficacy of a product, to compare two or more products, or to provide support for an advertising claim. In any of these cases, various statistical methods are available to analyze the data. In this paper, we will compare three methods of estimating the efficacy of a single antiperspirant product. Over the years Hill Top Research, Inc. has collected vast amounts of antiperspirant data, and this large base of historical information will be utilized for comparing the statistical methods of interest. This investigation suggests that one of the methods is preferable to the others. INTRODUCTION Evaluation of human perspiration evolved from simply viewing the amount of sweat produced during a set time period to collecting and weighing the actual sweat in the axillae. Collecting and weighing axillary perspiration through the use of absorbent pads has been used since the early 1950s and with minor modifications has remained fairly consistent for over 40 years. Analyses have also evolved over time. The analysis of the early visual data estimated reduction in the amount of perspiration seen. One form of analysis of the data collected with absorbent pads is done using treated axilla over untreated axilla ratios, adiusting the posttreatment ratios by a pretreatment ratio (1). As early as 1964 the use of analysis of variance to test for differences between treatments was employed (2). Adaptation of the analysis of variance (i.e., analysis of covariance, etc.) and other methods have also been used to analyze antiperspirant data. In this paper, we will consider experiments with the obiective of estimating the percent reduction obtained following use of an antiperspirant test product. That is, we will consider experiments designed to estimate the average percent reduction in perspiration that could be achieved by consumers using an antiperspirant. ANTIPERSPIRANT TERMINOLOGY Terminology that is typically used to define the various stages of an antiperspirant study is as follows: 13

14 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS CONDITIONING Each subject abstains from the use of axillary antiperspirants and uses only a deodorant product in the axillae for a period of 17 or more days (3) to effect washout of any previously used antiperspirant formulations. BASELINE Baseline sweat volumes and baseline axillary ratios are determined by sweat collections taken after the conditioning period, but before applying the antiperspirant test product. TREATMENT A series of supervised applications of the test formulation to the axillary areas is made during this period. The effect of treatment on axillary sweating is evaluated by sweat collections made at specific intervals following supervised applications. All antiperspirant studies have conditioning and treatment stages however, some do not utilize a baseline evaluation. Sweating of the test panelists is induced by having the panelists sit in a room maintained at 100øF --- 2øF and at a relative humidity of 35% --- 5%. During the first 40 minutes of the sweat stimulation period, the panelists hold unweighed pads of Webril (nonwoven cotton padding fabric) in their axillae. This preliminary warm-up period is followed by two successive 20-minute collection periods, during which the panelists hold weighed Webril pads in the axillae. These pads are placed in tightly capped polystyrene vials and weighed before and after use. STATISTICAL METHODS For each of the following methods, the source data for the statistical analysis consist of milligrams of sweat collected from right and left axillae of panelists who have partici- pated in antiperspirant studies. Treatments consist of one half of the panelists receiving a placebo, Sample X, on the right and the antiperspirant of interest, Sample Y, on the left, and one half of the panelists receiving the treatment in the opposite right-left order. The axilla that has been identified to receive the antiperspirant of interest, Sample Y, is designated as Axilla Y, and the other axilla as Axilla X. ADJUSTED RATIO METHOD For this method of analysis, pretreatment Y-to-X ratios are determined for each subject using baseline sweat collections. This ratio is calculated for each subject by Pretreatment ratio = Baseline milligrams of sweat from Axilla Y Baseline milligrams of sweat from Axilla X Posttreatment ratios are similarly calculated by