SURVEY OF COSMETICS FOR SEVEN INORGANIC ELEMENTS 139 Table XXII. Lotion with Sunscreen Summary (n = 5) Results (mg/kg) As Cd Cr Co Pb Hg Ni Maximum TR NF TR 0.045 0.10 NF TR Minimum TR NF TR NF NF NF NF Mean 0.12 NF 0.11 0.021 0.19 NF 0.036 Median 0.13 NF 0.10 0.019 NF NF 0.044 NF: Not found, or less than detection limit TR: Trace, or greater than detection limit, but less than quanti- tation level. stannous chloride through an online mixing system to reduce mercury to its elemental state. A gas–liquid separator removed the mercury vapor, which was deposited on to a gold bed through amalgamation. The gold bed was then heated to release the mercury vapor, which was carried by argon gas into the atomic fl uorescence spectrometer. Mercury was determined at 253.7 nm. Analytical solutions were analyzed relative to calibration standards using a dilution of 1/10 with diluting solution and 1% (v/v) bromine mono- chloride solution (to maintain mercury in Hg(II) state). Continuing calibration was verifi ed by analyzing a calibration blank and two QC solu- tions (secondary check solution and NIST 1641e reference standard) immediately after calibration and after every 15 analytical solutions. The QC solution recovery was between 90% and 110%. Absence of matrix effects was verifi ed by comparing one analytical solu- tion from each digestion batch with one that was diluted 1/20. RPDs for analytes in solu- tions at the two dilutions were less than 20%. Any analytical solutions with analyte levels above the highest standard were additionally diluted to determine the concentrations. RESULTS LOD/LOQ STUDY Ten replicates of reagent blanks fortifi ed with low level standards were analyzed for each element. The LODs, calculated as three times the standard deviations, and LOQs, calcu- lated as 10 times the standard deviations, are shown in Table II. Table XXIII. Mascara Summary (n = 10) Results (mg/kg) As Cd Cr Co Pb Hg Ni Maximum 0.35 NF 5.8 12 0.41 0.060 12 Minimum TR NF 0.26 0.15 NF NF 0.50 Mean 0.20 NF 1.8 2.8 0.13 0.0065 4.0 Median 0.17 NF 0.99 1.7 0.074 NF 2.3 NF: Not found, or less than detection limit TR: Trace, or greater than detection limit, but less than quanti- tation level.

JOURNAL OF COSMETIC SCIENCE 140 Table XXIV. Foundation Summary (n = 10) Results (mg/kg) As Cd Cr Co Pb Hg Ni Maximum 0.40 TR 5.0 2.5 1.2 TR 5.8 Minimum TR NF 0.46 0.16 0.16 NF 0.56 Mean 0.18 0.0072 2.3 0.77 0.64 NF 1.8 Median 0.14 NF 2.0 0.47 0.69 NF 1.2 NF: Not found, or less than detection limit TR: Trace, or greater than detection limit, but less than quanti- tation level. Table XXV. Body Powder Summary (n = 5) Results (mg/kg) As Cd Cr Co Pb Hg Ni Maximum 0.80 NF 3.9 1.6 1.5 NF 4.6 Minimum NF NF TR NF NF NF TR Mean 0.32 NF 1.6 0.82 0.48 NF 2.1 Median 0.29 NF 1.7 1.1 0.21 NF 1.6 NF: Not found, or less than detection limit TR: Trace, or greater than detection limit, but less than quanti- tation level. REFERENCE MATERIAL STUDY As a measure of accuracy, four reference materials were analyzed. Recovery limits were selected as 70%–130% for all certifi ed values less than 10 times the LOQ. For certifi ed values above 10 times the LOQ, recovery limits of 90%–110% were used. As a further test of ruggedness, recoveries were tested for three different weights of reference material: 0.25g, 0.50g, and 0.75g. There was a trend toward higher recoveries with lower weights for the custom oil standard for arsenic, cadmium, and possibly cobalt and nickel. These trends, however, were not deemed to be signifi cant with respect to variation in results. The recovery of mercury from its custom oil standard was also lower than for the other reference materials for an unknown reason. The recoveries for mercury ranged from 71% to 79% for 0.25g and 0.75g custom oil, respectively. The variation in recovery for differ- ent weights is not deemed to be signifi cant. Certifi ed values for the analytes in each refer- ence material and mean recoveries (from values obtained by analyzing the reference materials with each sample batch) are shown in Tables III and IV. The recoveries demon- strate good accuracy of the methods for all analytes. SUMMARY OF METHOD PERFORMANCE AND COSMETICS SURVEY There were no evident systematic errors. All calibration curves (y = ax + blank) had cor- relation coeffi cients of at least 0.997. All initial and CCV standards were within the control limits of 90%–110% recovery. All RPD for duplicates were less than 20% with