6. AOACSPIFANMethods-2018Awards

49

H ostetler : J ournal of AOAC I nternational V ol . 100, N o . 3, 2017  9

of the major cis isomers of α-carotene. Although this could potentially cause error in the β-carotene calculation, even in a sample with high α-carotene (Figure 2016.13D ), the cis - α-carotene/ cis -β-carotene peak accounted for only 5% of the total β-carotene peak area. To test whether the all- trans isomers of lutein and β-carotene were isomerized during the sample preparation, chromatograms from spike and recovery experiments ( n = 3) were used. Samples were spiked with all- trans carotenoid standards along with internal standard and carried through the preparation. No cis isomers of lutein were detected in the standard mixture, and none were detected in the spiked sample. In the β-carotene standard solution, cis isomers of β-carotene accounted for 3.4% of the total peak area in the standard mixture and 3.8% of the total β-carotene peak area in the spiked sample. This indicates that any isomerization of all- trans -lutein or β-carotene during the sample preparation is negligible. Linearity of the relative responses of analyte concentrations was measured using a five-point standard curve on 3 different days. Coefficients of determination, visual inspection, residuals, and relative errors of back-calculated concentrations were used for evaluation. Linearity of the internal standard was also tested. Regression lines for all- trans -lutein, all- trans -β-carotene, and apocarotenal are shown in Figure 4. Regression data for residuals and back-calculated concentrations are shown in Tables 1–3. The determination coefficients (R 2 ) for each curve were >0.999. The y -intercepts for all of the curves appeared insignificant; to test this assumption, sample calculations for all- trans -lutein and all- trans -β-carotene were performed by using both the y -intercept and forcing the y -intercept through zero. Two infant formulas were used: one with typical lutein and β-carotene concentrations and one with concentrations near the LOQ. The results (Tables 4 and 5) indicate that even for very low concentrations (3–4 μg/100 g) the difference between the two calculations was not more than 3%. Only when concentrations were near 1 μg/100 g did the calculations differ by as much as 9%. Based on these data, the y -intercept was forced through zero to simplify the calculations. In accordance with SMPR 2014.014, all data for infant formula and adult nutritionals are presented on a reconstituted basis (as is for RTF liquids, 25 g powder/225 g reconstituted weight for powder samples, or 1:1 by weight for liquid concentrates). The ranges for lutein and β-carotene (4–240 μg/100 mL) correspond to approximately 0.8–45 μg/100 g for samples prepared for the lowest sample concentrations. With dilutions specified in the method, the range can be extended to approximately 2250 μg/100 g. This range extends beyond that of 1–1300 μg/100 g specified in the SMPR. The LOD and LOQ were extrapolated from the S/N calculated in ChemStation software (Agilent Technologies, Santa Clara, CA) when measuring analyte concentrations of 1.4–1.7 μg/100 g in spiked NIST SRM 1849a. The LOD was calculated as (3 × measured concentration)/(S/N). The LOQ was calculated as (10 × measured concentration)/(S/N). Results from three different spiked samples were averaged. The Linearity LOD/LOQ

Figure 4. Linearity plot for (A) lutein, (B) β-carotene, and (C) apocarotenal.

determined LOQ for both lutein and β-carotene (Tables 6 and 7) meet the LOQ requirement of ≤1 μg/100 g in SMPR 2014.014.

Precision

Precision experiments were performed using the full SPIFAN sample kit, designed to represent current infant formulas and adult nutritional drinks on the market, in addition

Table 1. Residuals for the internal standard Apocarotenal concn, μg/100 mL Residual 203.5 4.119148067 101.8 –7.399430255 40.7 –3.177165248 20.4 1.435593088 4.07 2.222225757 2.04 2.79962859