228 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS Table IV Recovery of NDELA From Ethanolamine Derivative Matrices Matrix No. Mean of NDELA Specl- Found mens (ng/g) Recovery at Fort•fication Level* 50 ng/g 100 ng/g 200 ng/g 300 ng/g 400 ng/g 500 ng/g 600 ng/g 1,000 ng/g Overall DELALS 3 290 --+ 125 a DELALS 5 280 ñ 15 d TRELALS 2 46 -+ 25 f TRELALS 2 43 ñ 13 TRELA-CAP 3 48 ñ 3 a TRELA-CAP 3 ND, g, T, h T DLS + CD 1 910 Cocam•de MELA 4 ND Laumm•de MELA 2 ND Lam'am•de DELA 2 ND, 185 ñ 5 • Lmoleam•de 5 ND, ND, ND, DELA 102 ñ 18a, T __b __ -- 125 ñ 17 '• 115 ñ 19 • 125 --+ 19 • 120 ñ 17 ( -- 102 ñ 1 e -- -- 92 --- 98 --+ 6 a -- 104 ñ 4 c 106 ñ 10 e --- 110 -- -- 107 ñ 4 • 106 ñ 7' 104 -- 115 110 • 6 • -- -- -- 96 ñ 6 a 116 ñ 10 '• 1(/8 -+ 12 • -- 128 ñ 3 e -- -- 128 ñ 3 c -- -- -- 100 117 109 ñ 8 • -- 101 ñ 9 • -- -- -- 109 ñ 9 • 110 -- 98 .... 104 -- 6 • 110 ñ 10 e --- 95 -+ 8 e -- 107 ñ 1 • 104 -+ 11 • 81 ñ 11 • 79 ñ 1 d ..... 80 _+ 10 k a Mean of three determ•nanons ñ standard dewar•on (S D ) b No. data. c Mean of six determinations ñ standard dewauon (S D ) d Mean ofs•x determ•nauons (three specimens m duphcare) ñ standard dewauon (S.D) e Mean of two determinations ñ range. f Mean of four determinations (two specimens •n duph•ate) ñ standard dev•atxon (S D.) g Not detected (50 ng/g). h Trace(50 100 ng/g). • Mean of four determmanons ñ standard dev•auon (S.D.). I Mean of five determinations ñ standard dewat•on (S D.) k Mean of 10 determ•nauons ñ standard dewauon (S D ) * Recoveries based on NC-NDELA most convenient water removal method was judged to be vacuum desiccation. To aid in transfer of the dried sample and reduce irreversible adsorption to the glass vessel walls, 3 g silica gel (sufficient to make a thick slurry) were added to the sample prior to water removal. The resultant silica gel cake following desiccation was easily pulver- ized and added to the silica gel column for chromatographic elution of the NDELA as discussed previously. Selection of an internal recovery standard. The extraction, cleanup, and analysis process for NDELA in these matrices involves several steps which may affect the accuracy of the method by loss of analyte and/or erroneous volume measurements. Even though the methods are validated for many common cosmetic matrices, many variants which will most probably exhibit different recoveries may be encountered. Therefore, it is imper- ative that recovery be assessed for each sample type and preferably each individual sample. The most direct method for assessing recovery is with an internal recovery standard (IRS). The ideal IRS has the same chemical properties as the analyte so that, assuming the IRS is thoroughly incorporated into the sample, recoveries of the IRS and the analyte are identical. In addition, it must possess some independent measure- ment property so that the recovery may be determined. Clearly, the candidate IRS which best meets these criteria is a radiolabeled analog of the analyte in this case it was •4C-NDELA. The only detriment to this approach is that •4C-NDELA and unla- beled NDELA co-elute and respond to the TEA (and most other HPLC detectors) identically. Thus the concentration of the •4C-NDELA must be sufficient for scintil- lation counting without contributing significantly to the HPLC-TEA response. The specific activity of the •4C-NDELA used in this study was 10 mCi/mmol. 100% recovery at a spiking level of 26 ng per sample (about 260 pg per injection on the

NDELA DETERMINATION IN COSMETIC INGREDIENTS 229 HPLC) or 5.2 ng/1.0 ml counting aliquot would give about 1,000 disintegrations per minute or 800 counts per minute, assuming 80% counting efficiency. With a 10-min counting period, this level of NDELA has been found to give sufficiently precise results for the purpose of this procedure. The •4C-NDELA recovery for both the desiccation and column chromatography steps averaged 71 ___ 2% from four samples of amphoteric matrix and 80 --- 5% from eight samples of quaternium matrix. Va/idatio,. The results of the validation are shown in Table III. As can be seen from the table, the recoveries were quantitative and the methods were considered validated. MORPHOLINE MATRICES Method development. The morpholine matrix was found to be amenable to the simple dilution methods developed for the ethanolamine matrices. Not only did the morpho- line matrix samples readily dissolve in methanol/methylene chloride (4/96) but also little interferences were observed by HPLC-TEA using the Spherisorb CN column. This simple method was judged sufficient for the morpholine matrix. ACCURACY AND PRECISION The accuracy of the method may be estimated using the internal recovery standard. Assuming the •4C-NDELA and unlabeled NDELA have identical chemical behavior and therefore recoveries, the accuracy of the method depends only on the precision and accuracy of the •4C-NDELA measurement. Any systematic bias in the accuracy would have to be measured by an interlaboratory comparison of the method. Thus, the current best estimate of the method accuracy is 100%. As a corroboration of this conclusion, the mean of the •4C-NDELA-based validation data (Table III) is 107 -+ 12%. The overall mean of the validation data in Tables I through IV is 98.7 --- 22% for 93 determinations. Method precision may be measured using the estimate of the relative standard deviation obtained from the 93 validation experiments listed in Tables I through IV. While the precision varies slightly from matrix to matrix, the mean overall estimate of precision was calculated as _+ 22%. CONCLUSIONS Development of methods for NDELA in cosmetic ingredients is difficult sensitive, specific analytical techniques are required to reliably identify the NDELA at the low levels of interest to industry and government. Furthermore, the variety of matrices encountered requires a diverse set of sample preparation techniques. Research presented elsewhere discusses the first problem: development of HPLC~TEA methods for NDELA (18). This paper has presented a set of methods applicable to many, but not nearly all, cosmetic ingredient matrices. Where possible, simple methods are recommended (e.g., simple dilution). For more complex matrices, water removal and column cleanup tech- niques are prescribed.