205 / 363
S
alvati
et al
.:
J
ournal of
AOAC I
nternational
V
ol
.
99, N
o
.
3, 2016
7
Linearity
This method includes six working standards to bracket the
distribution of vitamin concentrations in SPIFAN II products.
Calibration curves were generated at the beginning and end
of each analysis as required by the method. Each standard in
the curve has its percent deviation calculated as the percent
difference between the calculated concentration and the true
concentration. Percent deviation of ±4% is typical for vitamins
B
1
, B
2
, B
3
, pyridoxamine, and pyridoxine; and percent deviation
of ±11% is typical for pyridoxal, which has lower response.
Good performance was observed (Table 2).
Accuracy
Accuracy was evaluated by over-spike recovery in the five
SPIFAN II placebos and three of select SPIFAN II products
(Table 3). The placebos were manufactured without fortification
of vitamins and minerals, but do contain some inherent vitamins
and minerals by contribution of the proteins, carbohydrates,
and fats. An additional three fortified SPIFAN II samples were
chosen for over-spike studies because they were different
enough from the placebos to warrant additional inquiry:
partially hydrolyzed, milk-based infant formula powder;
partially hydrolyzed, soy-based infant formula powder; and
soy-based infant formula powder. For over-spike recovery, each
matrix was spiked at both low and high levels corresponding
to approximately 50% and 200% of fortification, respectively.
Each spike level was performed with independent sample
preparation, and the experiment was repeated on three different
days for a total of
n
= 6 data points at each level in each matrix.
Recovery was calculated as the reported concentration divided
by the inherent contribution plus the amount spiked. All vitamin
forms required by the SMPRs were combined in the spiking
solution except thiamine triphosphate, which was not available
for purchase. Over-spike levels for each form were targeted to
mimic ratios previously reported in infant formulas and milk:
thiamine monophosphate and thiamine diphosphate were spiked
at 12.3% and 8.6% of total B
2
; riboflavin phosphate and flavin
adenine dinucleotide were spiked at 18.1% and 8.8% of total
B
2
; nicotinic acid was spiked at 7.2% of total B
3
; and pyridoxal
and pyridoxal-5′-phosphate were spiked at 4.9% and 4.3% and
pyridoxamine and pyridoxamine-5′-phosphate were spiked at
5.8% and 5.0% of total spiked B
6
. On an RTF concentration
basis, over-spikes were 2.60 and 21.0 μg/100 g of total
pyridoxal; 3.00 and 24.0 μg/100 g of total pyridoxamine; 22.5
and 180 μg/100 g of total pyridoxine; 31.5 and 250 μg/100 g of
total thiamine; 24.0 and 190 μg/100 g of total riboflavin; and
190 and 1500 μg/100 g of total B
3
. Good over-spike recovery
was demonstrated (Table 3).
Precision
Repeatability and intermediate precision were determined
from six independent preparations of all 14 products over
6 days. The experiments were performed by two analysts and
on one instrument. Repeatability and intermediate precision
are reported as %RSD in Tables 4 and 5. SPIFAN SMPRs for
repeatability and reproducibility are ≤5% and ≤10% RSD,
respectively.
Robustness
Method robustness was evaluated during development by
using three analysts and two instruments. The method was tested
over 6 days as well with independent preparations for each data
point, and accuracy was done over an additional three days
for each matrix. Data were collected over the course of about
8 weeks. Given these variables, precision and accuracy were
excellent suggesting good method robustness. Further, a review
of sample weights collected during sample preparation show that
the powder weight varied by up to 6%, the reconstitution weight
varied by up to 8%, and the liquid sample weight varied by up
to 9%. Given the demonstrated precision and accuracy, this
method shows good robustness toward sample size variation.
Within a run, there is notable signal suppression in some
matrixes. Suppression is most easily observed by noting the
absolute change in the internal standard intensity in samples
compared with standards. The degree of suppression is matrix-
and vitamin-dependent and ranged from negligible up to loss
of 50% of the signal. Ion suppression is not uncommon with
ESI, and necessitates the use of stable-isotope labeled internal
Table 2. Calibration curve % deviation from true concentration is reported at each calibration level
a
Standard
Overall (
n
= 12)
Thiamine Riboflavin
Niacin Nicotinic acid Pyridoxal
Pyridoxamine Pyridoxine
WS1
Recovery (%)
99.2
99.6
98.3
100.5
104.4
101.3
100.9
RSD (%)
3.5
6.6
5.9
7.3
15.4
7.2
2.3
WS2
Recovery (%)
100.4
99.9
100.8
99.1
97.9
100.3
98.8
RSD (%)
3.2
4.9
5.1
4.7
10.2
3.5
1.6
WS3
Recovery (%)
100.3
100.3
101.2
100.0
95.3
97.5
100.0
RSD (%)
2.1
3.7
3.2
2.2
10.7
4.6
1.3
WS4
Recovery (%)
100.5
100.6
100.5
100.4
104.3
99.8
100.4
RSD (%)
2.2
3.4
3.4
2.6
9.9
3.6
2.0
WS5
Recovery (%)
99.6
99.6
99.3
99.8
98.9
100.7
99.9
RSD (%)
2.4
1.9
2.5
1.8
10.9
4.0
1.4
WS6
Recovery (%)
100.1
100.1
100.2
100.0
100.1
99.8
100.0
RSD (%)
2.2
1.1
3.0
1.8
8.7
3.2
1.1
a
The reported value is averaged across 6 days and reported along with %RSD.
Candidates for 2016 Method of the Year
204