334 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS pyridine to give a final concentration of 0.1 g/L (100 ppm). Calibration standards were prepared from these stock solutions by dilution in methanol (MeOH). COSMETIC SAMPLE CLEANUP Approximately 12 g of cosmetic product were used for each analysis. The sample was split into two 6-g portions and each was accurately weighed into a 50-mL screw-cap tube (150 mm X 25 mm). A sufficient amount of NaCI was added to each tube to break the emulsion or inhibit formation of one. In the case of a viscous O/W hand cream, ca. 3 g per tube were sufficient. A 10% (w/v) ammonium sulphamate solution was prepared with deionized water and 9 mL of this solution were added to each tube to scavenge nitrite from the sample and prevent artifactual NNA formation. Ca. 0.5 mL of glacial acetic acid was pipetted into each tube, depending on the difficulty encoun- tered in breaking the emulsion (0.7 mL was used in the hand cream sample). After shaking by hand to dissolve the sample, 30 mL of 1,2-dichloroethane were added to each tube. The tubes were tightly capped and mixed using a vortex mixer for 2 to 5 min to ensure extraction of NDELA into the aqueous layer. At low pH, the nitrosamine primarily exists as the hydrochloride, thus favoring extraction into the aqueous phase. The tubes were centrifuged for ca. 10 min at 200 x g to completely separate the aqueous and organic layers. A disposable pipet was used to quantitatively transfer the top aqueous layer from each sample tube, combining them at the top of a single Extrelut QE © Kieselguhr column (#901020-1, EM Science, Cincinnati, OH). The aqueous sample layer was allowed to soak in for 3 min before extracting the NDELA with 60 mL of methyl ethyl ketone (MEK) in three successive aliquots of 20 mL each, waiting 5 min between the addition of each aliquot. The combined aliquots were passed through a SEP-PAK © silica car- tridge (#51900, Waters Associates, Milford, MA) into a 100-mL '1•. 24/40 round- bottom flask. The SEP-PAK © cartridge, containing strongly polar material from the sample, was discarded. The MEK was removed under water aspirator vacuum with a rotary evaporator, operating with a water bath .temperature of 40øC. The residue was taken up in 75 mL 1,2-dichloroethane, shaken vigorously with a vortex mixer, and then passed through a SEP-PAK © silica cartridge, retaining the NDELA on the cartridge. The NDELA was eluted with a 15-mL aliquot of ethanol into a 25-mL '• 24/40 round-bottom flask and the cartridge discarded. The ethanol was removed from the eluate by rotary evaporation at 40øC, and the flask containing the residual NDELA was stored in a desiccator until derivatization and analysis. This procedure is summarized in Figure 2. RESULTS OPTIMIZATION OF ELECTROLYTIC CONDUCTIVITY DETECTOR (ELCD) It has been demonstrated in other studies that reaction gas flow rate, conductivity solvent flow rate, and reactor furnace temperature can affect the response of an ELCD (38,40,41). It is generally believed that reaction gas flow rate is a variable only in that there must be a sufficient quantity to allow for complete reaction of the eluting nitrosa- mine. Typically, a flow rate of 50 mL/min has been shown to be adequate. A study was

ASSAY OF NDELA 3 3 5 6 g of sample weighed into 2 tubes with 3 g of NaCI, 0.5 ml glacial acetic acid, 30 ml 1,2-dichloroethane and 9 ml ammonium sulphamate solution Samples extracted with vortex mixer. Samples are centrifuged to separate phases. Aqueous layers combined on ExtrelutketoneQE•olethylmethyl NDELA eluted with 3 x 20 ml •M•K extract eluted through silica SEP-P© cartridge into round bottom flask to removestron polar material. MEK removed by rotary evaporation Residue dissolved in 1,2-dichloroethane j 1,2-dichloroethane eluted through silica SEP-PAK © cartridge NDELA retained on SEP-PAK © cartridge eluted with ethanol into round bottom flask Ethanol removed with rotary evaporation. Residue ready for derivatization and analysis Figure 2. Summary of cosmetic sample cleanup procedure. done to optimize the solvent flow rate, in which repeated 100-ng NDELA-TMS injec- tions were made at incremental flow rates and peak areas measured. The results (Figure 3) indicate that as the flow rate is reduced, the response increases. Venting the reaction furnace after sample injection causes a deflection of the detector baseline. As the flow rate is reduced, the detector baseline takes longer to settle after venting and becomes noisier. A conductivity solvent flow rate of ca. 160 p.L/min was selected as a good compromise between response, baseline settling time, and noise.