278 JOURNAL OF COSMETIC SCIENCE ensure comprehensive coverage of the brushed sample. An oscillation rate of 150 cycles per minute and a brushing force of 4.4 N was chosen to typify observed brushing practice (4). Cleaning performance was found to be relatively insensitive to brush head forces in the typical range. The force was applied to conventional 34-tuft fiat-trim 0.2-mm bristle nylon toothbrush heads (Mentadent P Professional ©, soft Elida Gibbs Dental Division, London, UK) via weights loaded onto vertical spindles mounted in linear ball bearings. Soil removal after 50, tOO, and 150 oscillations was monitored. A control paste containing 10% silica xerogel abrasive was included in each determination, done in duplicate. The test was deemed valid when the control paste performance remained within + 5% of its running mean value. Photometric measurements. The Minolta Chromameter © CR-300 (8-mm aperture) was used to measure the wh'iteness (L*) of the discs before soiling (clean), after soiling (soiled), and after cleaning (cleaned). Removal of the FT film was given by the equation: L* (cleaned)- L* (soiled) % Removal = X tOO L* (clean)- L* (soiled) CLINICAL STAIN-PREVENTION STUDIES Protocol. A panel of 45 adults (predominantly female, aged 18-50 years) was used to determine the stain-prevention properties of the test pastes in vivo. Particular consider- ation was given to the subjects' rights and personal convenience. Informed consent was gained for all participants. Four pastes were tested, blind, for six weeks each, end to end, in a fully randomized crossover design. At the beginning of each six-week period the labial aspect of the eight incisors of the subjects were scaled and polished by a dental hygienist. The subjects were given a standard soft-nylon fiat-trim toothbrush (Mentadent P Professional ©) and a supply of paste to use at home in their normal manner. Panelists complaining of unsightly stain levels on their teeth during the test periods were called in for examination and scoring by the clinicJan. Following a temporary cleanup, they were given the most abrasive paste to use for the remainder of the six-week period. This modification to the protocol was felt to be necessary in retaining the goodwill and compliance of the panelists over the whole 24-week trial period. It was exercised by only a small number of panelists using the least abrasive paste. Method of assessment. The degree of extrinsic stain after each six-week period was deter- mined by two assessors (J.E.W., T.M.) using substantially the method of Lobene (1968) (t 1). Thus Lobene's four-point intensity criteria were applied to the three regions of each tooth: gingival, interproximal, and body, viz., 0: no stain t: light stain (visible only under magnification) 2: moderate stain (visible to naked eye) 3: heavy stain The assessors were at no time directly aware of which paste each subject had been using during the trial. Agreement between, and reproducibility of, the assessors was excellent. Data. The proportion of subjects exhibiting visible stain (scores 2 or 3) on at least one site within each six-week period was determined for each paste. Mean total stain scores

DENTAL STAIN PREVENTION 279 associated with each paste were also recorded. However, owing to exercise of the pan- elist-withdrawal option during use of the 1ow-abrasivity paste, comparable total stain scores for all the pastes were not possible. Pastes. Four pastes were tested, representing a range of abrasivity from substantially zero to that typically found in regular toothpastes on the market. The abrasive performance of the pastes was determined by the new in vitro method at product concentrations of 33% by weight, slurried in water. The pastes were formulated to contain base ingredients typical of currently marketed "silica" formulations (sorbitol, sodium lauryl sulphate, and flavor). Thus the principal water-insoluble components were both thickening and abrasive-type amorphous silicas and opacifying levels of titanium dioxide. Of these, only the abrasive silicas exhibit any significant abrasive function. Experimental pastes were made containing zero, 2%, 4%, and 10% by weight of a typical commercially available abrasive xerogel silica (Gasil © 200TP, Crosfield Chemicals Ltd., Warrington, UK). This abrasive is representative of the efficient, high-strength, angular dental silicas of 5 to 10 microns average particle size used extensively in modern toothpastes. The unoccupied formulation space in the pastes containing less than 10% abrasive silica was filled with sorbitol and thickening silica in proportions necessary to maintain the viscosity and flavor of the 10% variant. RESULTS IN VITRO ABRASIVE EFFICIENCY The performance of the four pastes is shown graphically in Figure 1. Abrasive removal of the FT film from the polished HAP discs is plotted against the number of brush head oscillations applied. In this method it will be seen that the most efficient pastes reach a cleaning plateau at about 100 oscillations, and that the same order of effectiveness between pastes is ob- served at 50, 100, and 150 oscillations. Equally, rates of cleaning differentiate the pastes in the same order. A convenient simple expression of comparative abrasive performance was taken to be the percentage clean at 100 oscillations (FT-100 score). Under these conditions, mean differences of 5 percentage units are required for significance (p 0.01). Thus the in vitro performance of the pastes containing 10% and 4% abrasive silica was not substantially different, but a marked reduction was apparent for the 2% abrasive silica paste, with only slight abrasive cleaning provided by the zero-abrasive variant. No soil removal was observed after brushing with typical water-soluble components of pastes (e.g., humectant, surfactant), suggesting that the added titanium dioxide opacifier may be providing a small degree of abrasion. IN VIVO STAIN-PREVENTION EFFICIENCY In Table I below performance is expressed both as the percentage of the panel exhibiting visible stain (2,3 scores) on at least one site and as the mean total-stain score per panelist