W/O/W EMULSION STABILITY 341 SAMPLE TREATMENT Spectrophotometric methods. Our criteria for the method development were non-turbidity of the final solution, recovery, and feasibility. Non-turbidity was of particular importance to ensure the accuracy and precision of spectrophotometric measurements. Various treatments based on previous reports (13) were investigated with sodium chlo- ride solution and are detailed in Table I. The sodium chloride (4.32 g/l) solution is used to maintain the osmotic gradient. The main problem was to remove the oil residues from the aqueous sample. ß One-step procedures (A,B,C): The experiments carried out by one-step procedures were not feasible: filtration using 0.45-ptm cellulosic membranes was found unprac- tical due to clogging problems (method A). Methods B and C did not completely remove the oil residues. The centrifugation at 1000 rpm (Method C) was not suffi- cient to ensure the removal of the oil residues, while higher speeds were likely to provoke a disruption of the oil drops. ß Two-step procedures (D,E,F,G): Clarification was investigated with tetrahydrofuran and acetone (Method D). The best results were obtained with tetrahydrofuran (two volumes for one volume of diluted emulsion), according to other authors (13). Re- moval of the oil drops by extraction using chloroform or ether (Method E) led to a poor elimination of the oil phase. The solution remained turbid after this process. ß Three-step procedures (H,I): Methods H and I led to a limpid solution, but were lengthy and difficult. The best results (non-turbidity of the final solution, recovery, feasibility) was obtained with Method D. The experiments were continued with this method. Detailed conditions are described in the experimental section. HPLC method. The only sample treatment consists of a dilution of the multiple emulsion (1/1) (w/w) with the sodium chloride solution to reduce the sample viscosity and ensure reproducible injections (15). Table I Sample Treatments Methods Conditions Results A Filtration on 0.45-}xm cellulosic filter Obstruction of filter B Decantation: 1 hr Cloudy solution C Centrifugation: 1000 rpm for 30 min Cloudy solution D Decantation (1 hr) and clarification acetone 1:2 Cloudy solution tetrahydrofuran 2:3 Limpid solution Decantation (1 hr) and extraction chloroform 1:2 Cloudy solution ether 1:2 Cloudy solution Centrifugation (1000 rpm for 30 min) and filtration on Cloudy solution 0.45-}xm cellulosic filter Centrifugation (1000 rpm for 30 min) and clarification Cloudy solution (tetrahydrofuran 2:3) Decantation (1 hr), centrifugation (1000 rpm for 30 min), Limpid solution and clarification (tetrahydrofuran 2:3) Decantation (1 hr), filtration on 0.45-}xm cellulosic filter, Limpid solution and clarification (tetrahydrofuran 2:3). F G H I

342 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS DIHYDRALAZINE ANALYSIS From its physicochemical characteristics, dihydralazine has several interesting properties that allow its analysis. First, the benzene ring, combined with the two hydrazine functions, allows a quantification by UV spectrophotometry at 309.5 nm. Second, due to its two hydrazine functions, dihydralazine can react with ninhydrine to yield a orange-yellowish colored compound that can be detected by visible spectrophotometry at 488 nm. This will permit the analysis of dihydralazine in this external aqueous phase by direct UV spectrophotometry, visible spectrophotometry after derivatization, or RP-HPLC with UV detection. The experimental conditions of chromatographic analysis were previously studied (15). These methods were employed after a validation process. Specificity was examined for each technique from a tracerless multiple emulsion. Data on limit of detection, accu- racy, and precision (repeatability and reproducibility) are summarized in Table II. The marker recovery was determined by spiking the external aqueous phase with a known amount of dihydralazine. The recovery was 98% with the UV spectrophotometric method, 102% with the visible spectrophotometric method, and 99% with the chro- matographic method. The comparison of the three methods was based on several criteria: the method com- plexity, the overall processing time, the specificity, and the limit of detection. The chromatographic technique offered the best performances, though the others were suit- able for this purpose. APPEARANCE KINETIC OF THE TRACER IN THE EXTERNAL PHASE Two batches of the same emulsion were stored for one month at 40øC and three months at ambient temperature, 25øC. Each batch was then analyzed in triplicate by the three methods. All kinetic curves exhibited the same profile (Figure 1). The amount of tracer in the external aqueous phase showed a rapid increase followed by a plateau. Consequently, we used an exponential equation to model the dihydralazine release as a function of time. An exponential model conveniently expresses the initial rapid increase and the final ceiling of the release process. The equation in use, which corresponds to a first order kinetic, is: (Pt - Po) = A (1 - e -kt) (2) where (Pt - Po) is the difference between Pt, the percent of dihydralazine liberated in the external aqueous phase at time t, and Po, the percent of non-entrapped dihydrala- Table II Parameters of Validation UV Visible Validation parameter spectrometry spectrometry HPLC Limit of detection (g/l) 6.10 - 3 4.10 - 3 7.10 - 5 Reproducibility (CV%) 0.77 0.82 0.84 Repeatability (CV%) 0.92 1.0 ! 0.95 Accuracy (%) 0.60 0.26 0. ! 3