JOURNAL OF COSMETIC SCIENCE 488 cantilever-based sensors for the detection of the three retinoids, 9-cis retinol, 13-cis retinol, and all-trans retinol, in solution (20). This method utilized changes in the resonance fre- quency of the micro-cantilever sensor, due to the attachment of the analyte molecule to the micro-cantilever’s surface, to detect and quantify retinoid molecules. While these studies demonstrate utility for the analysis of retinoids in cosmetics, none of the methods involved the simultaneous analysis of a wide variety of cosmetic matrices, using real cosmetic samples, for all three retinoids—from the highly polar retinoic acid, to the less polar retinol, to the non-polar retinyl palmitate. Nor did these methods in- volve solid-phase extraction of the retinoid from the cosmetic matrix prior to analysis. Given the diversity of cosmetic products, the complexity of their formulations, and the many potential interfering compounds contained in them, purifi cation prior to HPLC analysis is critical. Therefore, we sought to develop and validate a new procedure suitable for use in a survey of consumer cosmetics for retinoic acid, retinol, and retinyl palmitate. In addition, a limited survey of cosmetic products labeled to contain retinoids was per- formed to demonstrate the applicability of this procedure. MATERIALS AND METHODS REAGENTS AND MATERIALS The following HPLC grade reagents were used: hexane and ethyl acetate (Burdick and Jackson, Muskegon, MI), isopropanol and ammonium acetate (Fisher Scientifi c Corpora- tion, Fair Lawn, NJ), acetic acid ( J. T. Baker, Phillipsburg, PA), and methanol and dichlo- romethane (Mallikrodt, Phillipsburg, N.J.). Butylated hydroxytoluene (BHT) was obtained from Sigma-Aldrich (St. Louis, MO). Retinoids for preparation of standards were obtained from Sigma-Aldrich and were stored at −20°C. Purities reported by Sigma- Aldrich were as follows: 13-cis-retinoic acid (98%), 9-cis-retinoic acid (98%), all-trans- retinoic acid (98%), 13-cis-retinol (90%), all-trans-retinol (95%), all-trans-retinaldehyde (98%), and all-trans-retinyl palmitate (99%). De-ionized water was prepared with a Milli-Q purifi cation system from Millipore (Billerica, MA). The extraction tube and adaptors were obtained from Supelco (Bellefonte, PA). The Symmetry C18 analytical column (250 mm × 4.6 mm with a 5-μm particle size) was obtained from Waters Corporation (Milford, MA). Celite 545 was obtained from Fisher Scientifi c (Fairlawn, NJ). SAMPLES OF COSMETIC PRODUCTS Twenty-nine samples were purchased via the Internet or from local stores and included a wide range of different types of skin care products with retinol or retinyl palmitate listed as ingredients. Samples were stored in their original containers and packaging at room temperature until opened for analysis. CALIBRATION STANDARDS Actinic glassware was used for preparation and storage of all solutions of standards and samples. Seven stock solutions were prepared in a solvent comprised of 1/3 (v/v) hexane,

DETERMINATION OF RETINOIDS IN COSMETICS 489 1/3 (v/v) isopropanol, 1/3 (v/v) ethyl acetate, and 0.1% BHT. Each stock solution con- tained 1 mg/ml of one of the following retinoids: 9-cis-retinoic acid, 13-cis-retinoic acid, all-trans-retinoic acid, 13-cis-retinol, all-trans-retinol, all-trans-retinaldehyde, or all-trans- retinyl palmitate. Working standard solutions were prepared from the stock solutions and were used for HPLC peak identifi cation and calibration of detector response. Each working standard contained equal concentrations of all seven retinoids appropriately di- luted in the hexane/isopropanol/ethyl acetate/BHT solvent (above). Ten working stan- dards containing the following concentrations of all seven retinoids were prepared: 0.0003, 0.001, 0.002, 0.003, 0.01, 0.02, 0.03, 0.1, 0.2, and 0.3 mg/ml. Three sets of calibration curves were used to cover the range of expected retinoid concentrations in the samples. One set of calibration curves was derived from data obtained by duplicate injec- tions of the standards having retinoid concentrations of 0.0003, 0.001, 0.002, and 0.003 mg/ml. Data from duplicate injections of the 0.003, 0.01, 0.02, and 0.03 mg/ml stan- dards were used to obtain a second calibration set of curves. The third set of calibration curves was obtained from data collected following duplicate injections of the 0.03, 0.1, 0.2, and 0.3 mg/ml standards. Each set of calibration curves was obtained using a linear regression analysis of peak areas versus standard concentrations for each of the seven reti- noids. Retinoids isolated from samples were identifi ed by comparing their retention times and UV absorbance maxima with those of standards, and then were quantifi ed by using the appropriate standard calibration curve for each retinoid. When not in use, stan- dards and sample extracts were tightly sealed, stored at 2°C, and protected from light they were stable for up to three months. SAMPLE EXTRACTION Approximately 300 mg of each sample was weighed into a 40-ml beaker, mixed thor- oughly with about 1.7 g of Celite, and then transferred to a 10-ml extraction tube. The sample/Celite mixture was covered with a fi lter disk and compacted. The extraction tube was eluted into a 10-ml volumetric fl ask with enough extraction solvent (1/3 (v/v) hexane, 1/3 (v/v) isopropanol, 1/3 (v/v) ethyl acetate, and 0.1% BHT) to fi ll the volumetric fl ask to the mark. At the end of the extraction period, the eluate was thoroughly mixed and a 20-μl aliquot was immediately injected into the HPLC. Duplicate injections were per- formed for all samples. During preparation and analysis, exposure of standards and extracted samples to air, light, and heat was minimized to prevent oxidation or decompo- sition. When not in use, standards and sample extracts were tightly sealed, stored at 2°C, and protected from light. HPLC METHOD Chromatographic analyses were carried out with an Agilent 1100 series HPLC chromato- graph, equipped with quaternary pumps, a vacuum degasser, an auto-injector, a variable- wavelength diode array UV-visible absorbance detector, and a personal computer with chromatographic and spectrographic software. The mobile phase was degassed with an in-line degasser. Chromatographic separation was achieved using a Symmetry C18 analyti- cal column (250 mm × 4.6 mm with a 5-μm particle size). The column was eluted at 1 ml/min using a gradient starting at 25% solvent A (0.4 M ammonium acetate/1.0% acetic acid pH 5.3 buffer) and 75% solvent B (methanol),