AOAC Methods for Review in Codex STAN 234_11-2018

AOAC Official Methods Listed in CXS 234 for Milk and Milk Products

84

expected for a sample in which 4 times the amount of calcium phosphate could settle to the bottom of the container. The poor RSD R also may be due to greater can-to-can variation. This is indicated by the high RSD R for P for sample 6, which increases to 11 % from < 4 % for other test samples. The IUPAC protocol was used to identify outliers. Although more than the required 8 laboratories began the collaborative study, only 8 laboratories, including the author’s, completed the study. In Table 9, which shows results for determination of P, only 7 laboratories reported results. For whey powder sam- ple 3, removal of the Grubb’s outlier would have increased the outlier removal rate above the standard 2/9 laboratories. In ac- cordance with harmonization guidelines, the outlier was in- cluded in the reproducibility estimate. Table 9 also shows esti- mates without the outlier. Recoveries from spiked sample are listed in Table 11. Most recoveries are within the 90–110 % range, and results indicate the method is accurate for soy-based, whey-based, and enteral formulae. For all elements excluding Ca, the HORRAT ratio was < 1 for most test samples, indicating that between-labora- tory variation was slightly less than what would be expected for analytes at these concentrations. As shown in Table 10, repro- ducibilty estimates are close to those of the original study. In the case of liquid formula, it may be essential to include direc- tions to shake the container before weighing to ensure complete mixing, as stated in AOAC Official Method 984.27 , Calcium, Copper, Iron, Magnesium, Manganese, Phosphorus, Potas- sium, Sodium, and Zinc in Infant Formula, Inductively Cou- pled Plasma Emission Spectroscopic Method.

Test sample density should be ca 1.03 g/mL for RTF formula. Ref.: J. Assoc. Off. Anal. Chem. 69 , 777(1986); J. AOAC Int . 80 , 834–844(1997)

Results and Discussion

Results for Ca, Mg, Zn, Cu, Fe, Mn, K, Na, and P are sum- marized in Tables 1–9, respectively. RTF liquid soy formulae and soy powders with similar concentrations of analyte were pooled together as Youden “closely matched pairs” for esti- mates of within-laboratory variability (RSD r ). Despite a differ- ent technique in estimating repeatability and reproducibility (RSD R ) from that used in the original study, estimates of these parameters can be compared. Use of results from different-day analysis on the same RTF milk-based formulae in the original study produced estimates of repeatability and reproducibility close to those determined in this study. Averaged estimates of repeatability and reproducibility from the original collaborative study are listed with the averaged estimates from this study in Table 10. Exceptions are for Ca and P, which had twice the re- peatability and reproducibility estimates as in the original study. Calcium results with the highest variation had the poorest HORRAT ratios (ratio of relative standard deviation among laboratories/relative standard deviation expected, extrapolated from previous studies for analytes at those particular levels). The resulting relative standard deviations estimating between- laboratory variation were quite high, and for 2 test samples the HORRAT ratio was more than twice what would be expected at that concentration of analyte. However, no outlier among the Ca results was identified by the Cochran and Grubb’s tests for outliers. The sum ranking test proposed by Youden and Steiner (6) identifies laboratories with consistently high or consistently low results for a significant number of the test samples ana- lyzed. With the sum ranking test applied, 2 laboratories would rank significantly low and one laboratory would rank signifi- cantly high. Eliminating the high ranking laboratory results and the low ranking laboratory results for each test sample ( see bot- tom of Table 1) results in relative standard deviations for Ca between laboratories that are closer to the average reproduci- bility estimate of the original study as listed in Table 10. One collaborative study for Ca, Mg, and P in cheese (7) identified the potential for Ca contamination as a leading factor in between-laboratory variation. In this study, results from 2 laboratories ranked consistently low compared with those re- ported by other laboratories. No blanks with significant levels of Ca were reported, indicating that Ca contamination was not a problem. One of the low-ranking laboratories weighed a larger amount of test portion and exceeded the 50 g limit recom- mended by the method for sample 6. The other low-ranking laboratory used different amounts of sample for determining minerals by Method 986.24 . This method specifies that liquid formula must be adequately mixed before each weighing to prevent calciumphosphate from settling. The poorest RSD R for Ca was for the 32 fl. oz. can, sample 6. This variation would be

Recommendation

On the basis of the results of this study, it is recommended that AOAC Official Methods 985.35 and 986.24 be extended to include all types of infant formulae and enteral products.

Acknowledgments

I thank the following collaborators: Robert Allen, Covance Laboratories, Madison, WI Anjum Choudhry, U.S. Food and Drug Administration, Atlanta Center for Nutrient Analysis, Atlanta, GA Randy Fleener, Silliker Laboratories of Illinois, Chicago Heights, IL Nancy Miller-Ihli and F. Ella Greene, U.S. Department of Agriculture, Nutrient Composition Laboratory, Beltsville, MD Carol Johnson, Lancaster Laboratories, Lancaster, PA Rochelle Powell, Woodson-Tenent Laboratories, Inc., Memphis, TN Wai-Leung Yip, U.S. Food and Drug Administration, North Eastern Regional Laboratory, Brooklyn, NY I also thank Richard H. Albert and Richard F. Newell, Divi- sion of Mathematics, Office of Scientific Support, U.S. Food and Drug Administration, for statistical analysis of the data. I also acknowledge the support of Martin P. Bueno, former Gen- eral Referee for Infant Formula and Medical Diets.

10/9/2018