AOAC Methods in Codex STAN 234 (Preliminary Methods Review)

BARBANO ET AL.: J. ASSOC. OFF. ANi\l. CHEM (VOL. 73, NO. 6. 1990)

852

there were 11 complete sets of results from 10 different laboratories. A )-digit sample coding system was used. A computer program was prepared to collect data and to translate the codes to match blind duplicate test results, as well as to calculate AOAC statistical parameters and to conduct outli– er tests (25, 26). Special data forms were sent to each labora– tory to minimize the probability ofsample mix-up. There was 1 data form for each set of 18 milk samples. The form had all the sample numbers (3-digit code) printed on it in the order that the analyst was to test the samples. There was a form for the analyst to write comments about each individual sample, and a questionnaire (to be completed by the analyst) for each testing method. Information such as sample arrival time and actual testing conditions during the analysis was collected to help ensure that the analysts followed all the details of the procedures. Six: oz Whirl·Pak bags were coded with sample identifica– tion numbers. Raw milks (S L) were collected from 9 differ– ent farms on a Monday morning (day I) and transported on ice to the central laboratory, On Tuesday (day 2), milk samples were cold-split in the central laboratory: the milk from each farm was mixed, poured into 1 large plastic con– tainer, and agitated continuously with a motor driven stirrer while the milk was drawn from a spout (at the bottom of the container) directly into coded sample bags. Forty-six. or more 6 oz Whirl-Pak bags of milk were prepared (80 mL/bag). Samples were refrigerated immediately after splitting. Sam– ple splitting uniformity was verified immediately after split· ting with a Dairylab 2 infrared milk analyzer [16.083, 14th Ed.( I) J by checking the fat and protei.n test of the (irsl, middle, and last sample bag. On Tuesday afternoon, when milk from all 9 farms had been split, the samples were put into appropriate groupings, packed in ice, placed in insulated shipping containers, and sent by overnight air delivery to participating laboratories. Also, a clean, dry, 250 mL wide-moutb, screw-cap plastic container was sent to each laboratory so that the analyst could ship 250 ml of 0, lOOON HCI titrant (a representative sample from the acid used to titrate the distillates) to the central laboratory for analysis. On Wednesday (day 3), sam– ples arrived at the laboratories and testing was initiated as soon as possible. Arrival temperature of the milks was always ~4°C. All testing was completed by Monday (day 8). Test results and completed questionnaires arrived at the data analysis laboratory by Wednesday (day 10). AU data were summariz.ed and returned to individual testing laboratories by day 16. Collaborators verified that the data compiled for their laboratory were correct. V•rlllcaflon of MIik Sampl• Quality A set ofmilk samples was tested for somatic cell count with a Foss-o-Matic electronic somatic cell counter (27), which was calibrated by direct microscopic somatic cell count (28). This set ofsamples had been exposed to all the same handling and shipping conditions as the samples that were tested by the Kjeldabl method, Collaborat/11e Study Trllll• For the first collaborative study (11 sets of samples, 9 milks in blind duplicate in each) conducted in Septembel' 1988, the methods used were as stated in the methods section except analysts were requested to digest samples for I h after clearing. Based on the results of the first trial, a second

the final digestion temperature) was effective in controlling sample foaming during digestion. Laboratories participating in the study preferred this approach to control foaming. The fourth phase of preliminary work found that titanium oxide did not improve test results or substantially decrease digestion time. Kane ( I0) found that titanium oxide im– proved test results and decreased testing time on animal feeds. In this work, he also used a 0.835 ratio of potassium sulfate to sulfuric acid (instead of 0.6), which may explain the shorter testing time required for complete digestion. Higher ratios of K 2 S04 to H 1 S0 4 will give a faster digestion, but can result in excessive acid loss during digestion, crystal– li1.ation of tJ10 digest on cooling, and low test results ( 15). At a salt-acid ratio of 0.6, the mean test results for milk were 3.08% protein without titanium and 3.08% protein with tita– nium using the same time from start of digest.ion to clear and 60-min boil after clearing. With a 45-min boil after clear, lhc results were 3.08% protein without titanium and 3.07% pro– tci11 with added titanium. At a salt-acid ratio of 0.85, the mean test results for milk were 3.09% protein without titani– um and 3.09% protein with titanium using the same time from start of digestion to clear and 60-min boil after clear. With a 45-min boil after clear the results were J.09% without titanium and 3.08% with titanium. The digests with the high– er salt- acid ratio did not crystallize after cooling, but were extremely viscous and hard to mix with the water prior to distillation. With the higher salt- acid ratio, peak nitrogen recovery may be reached slightly sooner during the boiling period after clearing. There were no significant advantages to the use of titanium dioxide or a higher sail-acid ratio for determination of total nitrogen content of milk samples by Kjelda.hl analysis. The fifth phase of the preliminary work with 9 laboratories in July 1988 indicated that nitrogen ree-0veries in all labora– tories were acceptable. In a meeting attended by the Asso– c.1ate Referee and laboratory representatives in August I 988, it was suggested that hydrogen peroxide be eliminated from the digestion and that sample foaming be controlled in the block digestors by starting at a lower initial digestion tem– perature. Also, it was requested that 2 blanks instead of I blank be run by each l~boratory in the collaborative study, because l erroneous blank value would bias all the results from a laboratory. Finally, it was suggested that indicator be added to the boric acid receiving solution when the boric acid is first prepared. This "batch" addition of indicator would eliminate variation in amount of indicator between different samples and make the titration end color more consistent among samples. After completion of prelimfoary work and prior to the beginning of the collaborative study, complete and detailed descriptions of the final testing methods were developed and sent to all laboratories. Collaboratlve Study For the collaborative study, each of 10 laboratories re– ceived a set of 18 raw milk samples (9 pairs of blind dupli– cates) . Each pair of blind duplicates represented milk from I farm. Sample coding was designed such that all samples had different identification numbers in all laboratories. Sample testing order was randomized between laboratories. Ten lab– oratories participated and one of these laboratories conduct– ed 2 sets of analyses, one with traditional equipment and one with block digc.stor equipment (2 different sets of milk with different sample coding were sent to this laboratory). Thus,

Codex Trial Method Review