385 / 471
P a g e
|
4
collection of other laboratories’ data for the Certificate, as the Certified value is closer to the free
carnitine than to the total carnitine.
Tables 3a and 3b, Tabs 6&7, show the Repeatability statistics for Total and Free determinations,
respectively. These tabs also include Intermediate Precision data. In our laboratory, our SLV was carried
out as follows. On Day 1, Analyst 1 prepared duplicate free and duplicate total preparations of each
matrix and then split them into two portions. Half were put on Instrument 1 (a Waters TQD) and half
were put on Instrument 2 (a Waters TQMS). The samples were analyzed overnight. On Day 2, Analyst 2
repeated this procedure, and so on, until each Analyst had collected 3 Days of data. Thus, there are 6
days of independent preparations in this protocol, but additional information was collected as far as
how two different instruments’ results agreed (from the split samples, which were not independent).
The data shown in Tables 3a and 3b represent Days 1-3 from Instrument 1, and Days 4-6 from
Instrument 2. This kept all data independent while maximizing the sources of variance from different
days, analysts, and instruments. From the data on duplicates collected each day, ANOVA extracted the
repeatability data that is shown in Tables 3a and 3b, while the overall RSD for the six days represents the
intermediate precision, which is also listed. The Repeatability RSD for choline and carnitine were all in
the 2% range or lower, easily meeting the SMPR requirements of 5% for choline and 8.0% for carnitine.
The intermediate precision RSDs were mostly in the 2-3% range, which is a good indication that the
method will meet the required Reproducibility of 10% for choline and 15% for carnitine. Note that the
free choline level in Adult Milk Powder (0.36 mg/100g RTF, Table 3b) appears to be right around the LOQ
because the short term precision is fine, but the intermediate precision begins to degrade. This level is
about 10x lower than the required LOQ of 2 mg/100g from the SMPR, a level at which we obtained good
recovery (Table 5b).
Table 4, Reproducibility, Tab 8, is empty because these data have not been collected yet. They would be
collected during a MLT study, if the method is selected to proceed to such.
Table 5a (Tab 9), Linearity, gives some information about the calibration standards, sample sizes,
dilution factors, and the curve linearity. To demonstrate linearity, the normal calibration curve was
made, and then calibration standards were run as samples against this curve to determine the
calibration bias (residuals) at each level. The average recovery over 3 days for carnitine at the 3.4 µg/L
level (50% of the normal WS1 level) was 101.2%; for choline at the WS1 level of 30 ppb, it was 100.3%.
These residuals at low levels reflect the excellent linearity of the curve (usually r = 0.9999) and the low
blank values.
Table 5b, Tab 10, shows the results of spiking the Adult Milk Powder at a level that corresponds to the
LOQ required by the SMPRs of choline and carnitine. The spike recovery of choline (96.9%, 0.9% RSD for
seven replicates in one day) and carnitine (105.9%, 1.4 %RSD) indicate no problem with determining
these low amounts in a product matrix. The Adult Milk Powder product had virtually no inherent
carnitine and was therefore a good placebo for it. As shown in table, after spiking this product measured
0.17 mg/100g with a 1.5% RSD, indicating that there was no problem with the determination at the
required LOQ.
Candidates for 2016 Method of the Year
323