424 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO.

2, 2002

FOOD COMPOSITION AND ADDITIVES

Determination of Vitamins A (Retinol) and E (alpha-Tocopherol)

in Foods by Liquid Chromatography: Collaborative Study
JONATHAN W. DEVRIES and KARLENE R. SILVERA
Medallion Laboratories, 9000 Plymouth Ave North, Minneapolis, MN 55427

Collaborators: S. Al-Hasani; J. Alfiere; C. Berge; C. Boerner; S. Cardozo; M. Chettiar; K. Dupont; K. Gustafson;

E. Hanson; A. Kazeminy; D. Krueger; R. Mazal; P. Meland; B. Mioc; L. Oehrl; E. Vinski; D. Willis; B. Wittrig

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

A collaborative study was conducted for the deter- an AOAC Official MethodSM or as an approved method of the
mination of vitamins A and E. Existing AOAC liquid American Association of Cereal Chemists has not been
chromatographic (LC) methods are suited for spe- achieved. No modern methods exist for the simultaneous de-
cific vitamins A and E analytical applications. This termination of vitamins A and E by LC for all food categories.
method differs from existing methods in that it can With the need for simplicity and desire for a single flask
be used to assay samples in all 9 sectors of the work-up for more than one analyte, a method that has been
food matrix. Standards and test samples are successfully used routinely for nearly 25 years in the authors’
saponified in basic ethanol–water solution, neutral- laboratories was submitted to collaborative study. Rugged-
ized, and diluted, converting fats to fatty acids and ness testing of the method was performed as part of its initial
retinol esters and tocopherol esters to retinol and development and improvement (2–4).
tocopherol, respectively. Retinol and alpha- The concentration of vitamins A and E in foods and supple-
tocopherol are quantitated on separate LC sys- ments is usually given in international units (IU). The scien-
tems, using UV detection at 313 or 328 nm for tific literature often uses the unit, retinol equivalents (RE) to
retinol, and fluorescence detection (excitation account for the fact that different vitamers of vitamin A differ
290 nm, emission 330 nm) for alpha-tocopherol. Vi- in activity. Analysts determine the quantities of the vitamers
tamin concentrations are calculated by comparison themselves in mass units such as µg/g or mg/g, units which are
of the peak heights or peak areas of vitamins in test readily converted to the desired bioactive units.
samples with those of standards.
Collaborative Study
itamin A is extremely important for health: eye func-

V tion, antioxidant activity, the immune system, cell

growth, fetal development/reproduction, hormone
function, maintenance of skin condition, bone and teeth for-
Twelve food samples were selected on the basis of their fat,
protein, and carbohydrate content. The test samples were cho-
sen to match the ratios expressed in the AOAC food analysis
triangle (5) to cover the breadth of macro food properties. One
mation, and neurotransmitter function. Vitamin E is also im-
portant for health: muscular function, reproductive function, test sample representative of each of the 9 food sectors in the
and antioxidant activity. A difficult but important task of ana- triangle was selected. In addition, an extra test sample was se-
lytical laboratories is the accurate determination of vitamins A lected from each of the sectors 5–7 because most foods con-
and E in foods. These determinations have traditionally been sumed lie in these sectors. Judicious selection provided test
made using AOAC Methods 960.45, 974.29, 971.30, 975.43 samples with a wide range of vitamins A and E content. Eight
and/or 948.26 (1). In recent years, liquid chromatographic of the foods were spiked with known levels of retinol
(LC) technology has been applied and successfully adopted palmitate and alpha-tocopherol acetate so that typical recover-
for a number of specific vitamins A and E analytical applica- ies to be expected from this method could be determined. The
tions. Broad application of LC to the analysis of vitamins A test samples and their respective protein, fat, and carbohydrate
and E in foods leading to adoption of the technology as part of content are listed in Table 1.
All test samples were thoroughly homogenized in a
blender. Spikes were added as appropriate, and the spiked
Submitted for publication November 2001. samples were homogenized a second time. Subsamples of
The recommendation was approved by the Methods Committee on Food
Nutrition as First Action. See “Official Methods Program Actions,” (2001) about 20 g were packaged in glass containers, sealed, coded,
Inside Laboratory Management, November/December issue. and held at or below –20°C until shipment. Test samples were
Vitamin A was approved for Official First Action. Vitamin E was not removed from frozen storage just before shipment, shipped to
accepted for Official First Action.
Corresponding author’s e-mail: [email protected]. collaborators at ambient temperature overnight, with instruc-
tions to place in frozen storage until the time of analysis. All
 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002 425

Table 1. Foods selected for collaborative study

Food matrix sector Test samples Protein, % Fat, % Carbohydrate, %

1 Margarine/butter (50/50 mix) 0 100 0

2 Chicken gravy (canned) 6 42 52
3 Cheese sauce 19 60 21
4 Whole egg powder 42 48 10
5 Multigrain cereal 17 6 77
5 Corn cereal 9 0 91
6 Infant formula (powdered) 24 17 59
6 NIST SRM 1846 24 17 59
6 Baked beans with franks (60/40, w/w) 23 26 51

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

7 Dried milk 41 1 58
7 Full fat soy flour 42 23 35
8 Cottage cheese 66 20 14
9 Canned tuna in oil 86 14 0

collaborators were provided with one “check sample” to as- 328 nm for retinol, and fluorescent detection (excitation 290 nm,
sess their performance before beginning the assays. emission 330 nm) for alpha-tocopherol. Vitamin concentrations
are calculated by comparison of peak heights or peak areas of vi-
AOAC Official Method 2001.13 tamins in test samples with those of standards.
Determination of Vitamins A and E in Foods
Liquid Chromatography B. Apparatus and Materials
First Action 2001
(Vitamin E was not accepted for Official First Action) Note: Two separate simple isocratic LC systems can be
(Applicable for the determination of retinol from 0.15 µg/g used to perform the determination steps of this method for vi-
to 1 g/g and for the determination of alpha-tocopherol from tamins A and E simultaneously. Alternatively, the retinol and
0.001 mg/g to 1 g/g in foods.) alpha-tocopherol can be analyzed sequentially on a single LC
system since the alpha-tocopherol is sufficiently stable in the
Caution: Potassium hydroxide is extremely caustic. This diluted saponification solution to be stored while vitamin A is
chemical can cause severe burns. Protect skin and being run.
eyes while performing this method. This method (a) HPLC system.—(1) Pump.—A high pressure pump
involves the use of flammable liquids. Perform operating continuously at 1.0 to 2.0 mL/min with a flow preci-
behind a barrier when using hot water, steam, or sion of ±1% or better. (2) Injector.—A manual injector or
an electric heating mantle. Use an effective fume autosampling injector with a 20 µL fixed loop having a typical
removal device to remove flammable vapors as sampling precision of ±0.25% or better. (3) Chromatography
produced. Leave ample headroom in flask and columns.—(a) For retinol (vitamin A).—Reversed-phase
add boiling chips before heating is begun. Place C18, 10 µ (4.6 × 250 mm) capable of separating cis and trans
all controls, unless vapor sealed, outside of vapor isomers of retinol with a resolution of 1.0 or greater. Cis
area. This method utilizes toxic chemicals. Use retinol typically elutes prior to trans retinol on columns pro-
an effective fume removal device to remove va- viding effective separation. (b) For alpha-tocopherol (vita-
pors as produced. min E).—Reversed-phase C8, 10 µ (4.6 × 250 mm) capable of
separating alpha- and beta- or alpha- and gamma-tocopherol
See Tables 2001.13A–D for the results of the inter- and with resolution of 1.5 or better. The tocopherols will typically
intralaboratory studies supporting acceptance of the method. elute in the order of delta, gamma, and beta (beta and gamma
are rarely well separated on reversed-phase columns), then al-
A. Principle
pha. It is important to achieve adequate resolution between al-
Standards and test samples are saponified in basic etha- pha and the other tocopherols; otherwise low levels of al-
nol–water solution, neutralized, and diluted, converting fats to pha-tocopherol will be masked by high levels of gamma- and
fatty acids and retinol esters and tocopherol esters to retinol and beta-tocopherol in products with high levels of gamma- and
tocopherol, respectively. Retinol and alpha-tocopherol are beta-tocopherol (i.e., oleomargarines produced from soy oil).
quantitated on separate LC systems, using UV detection at 313 or (4) Detectors.—(a) Vitamin A.—Photometric detector mon-
426 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002

Table 2001.13A. Interlaboratory study results for the determination of vitamin A in foods by LC
Food matrix
sector No. Matrix Labsa(b) Mean, µg/100 g Rec., % sR RSDR, % R HORRAT

1 Margarine/butter (50/50 mix) 12(0) 857.08 68.36 7.98 191.41 0.69

1 Margarine/butter (50/50 mix; spiked) 10(2) 1457.20 106.6 89.90 6.17 251.72 0.58
2 Chicken gravy (canned; spiked) 9(1) 131.13 97.1 19.17 14.62 53.68 0.95
3 Cheese sauce (spiked) 11(0) 271.45 103.7 34.96 12.88 97.89 0.94
4 Whole egg powder 10(0) 161.98 62.48 38.57 174.94 2.59
4 Whole egg powder (spiked) 12(0) 426.12 91.7 205.71 48.28 575.99 3.75
5 Multigrain cereal 12(0) 634.38 114.85 18.10 321.58 1.49
5 Corn cereal 12(1) 945.52 86.78 9.18 242.98 0.80
5 Corn cereal (spiked) 12(1) 1395.91 108.5 76.11 5.45 213.11 0.51

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

6 Infant formula (powdered) 12(0) 584.15 95.98 16.43 268.74 1.34
6 NIST SRM 1846 12(0) 464.28 79.5 49.06 10.57 137.37 0.83
7 Dried nonfat milk 12(0) 817.27 102.94 12.60 288.23 1.08
7 Dried nonfat milk (spiked) 12(0) 1708.25 125.5 195.68 11.46 547.90 1.10
8 Cottage cheese 8(1) 46.02 8.19 17.80 22.93 0.99
8 Cottage cheese (spiked) 10(2) 411.03 98.9 69.29 16.86 194.01 1.30
9 Canned tuna in oil (spiked) 11(1) 262.36 89.5 56.11 21.39 157.11 1.55
Avg. 100.1 ± 13.2
a(b)
Number of laboratories where a = number of laboratories retained after outliers removed and b = number of outlier laboratories.

itoring absorbance at 328 nm. (Alternatively a wavelength of (d) Reflux condensers.—With adapters (if necessary) to
313 nm can be used.) (b) Vitamin E.—Fluorescence detector, attach 125 mL low actinic Erlenmeyer flasks and nitrogen
excitation 290 nm, emission 330 nm. (5) Recorder, intergrator, lines.
or data collection system.—Compatible with detectors used. (e) Volumetric flasks.—Low actinic 100 and 10 mL.
(b) Erlenmeyer flasks.—Low actinic 125 mL with neck (f) Nitrogen blanket apparatus.—A supply of nitrogen
adapted for connecting reflux condenser. gas with appropriate tubing and connectors to provide a con-
(c) Hot plate.—With sufficient heating surface area to stant nitrogen atmosphere blanket in the reflux apparatus dur-
handle multiple reflux apparatus setups preferred. ing saponification.

Table 2001.13B. Interlaboratory study results for the determination of vitamin A in foods by LC (Youden pair
statistical treatment)
Mean,
Youden pairs Labsa(b) µg/100 g sr RSDr, % sR RSDR, % r R HORRAT

Dried nonfat milk (spiked) and 10(2) 1595.1 107.29 6.73 112.22 7.04 300.41 314.22 0.67
margarine/butter (50/50 mix; spiked)
Corn cereal and corn cereal (spiked) 10(2) 1173.3 48.33 4.12 85.49 7.29 135.32 239.37 0.66
Margarine/butter (50/50 mix) and dried 12(0) 837.18 66.23 7.91 87.38 10.44 185.44 244.66 0.90
nonfat milk
Multigrain cereal and infant formula 12(0) 609.27 87.42 14.35 105.83 17.37 244.78 296.32 1.42
(powdered)
Cottage cheese (spiked) and NIST 10(2) 436.18 60.4 13.85 61.23 14.04 169.12 171.44 1.10
SRM 1846
Canned tuna in oil (spiked) and cheese 11(1) 269.25 39.14 14.54 45.07 16.74 109.59 126.20 1.21
sauce (spiked)
Chicken gravy (canned; spiked) and 8(1) 137.94 22.19 16.09 36.58 26.52 62.13 102.42 1.74
whole egg powder
a(b)
Number of laboratories where a = number of laboratories retained after outliers removed and b = number of outlier laboratories.
 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002 427

Table 2001.13C. Interlaboratory study results for determination of vitamin E in foods by liquid chromatography
Food matrix Mean,
sector Matrix Labsa mg/100 g sR RSDR, % R HORRAT Rec., %

1 Margarine/butter (50/50 mix) 7(0) 3.75 0.68 18.21 1.90 1.96

1 Margarine/butter (50/50 mix; spiked) 8(0) 13.77 2.07 15.05 5.80 1.97 103.03
2 Chicken gravy (canned; spiked) 8(0) 4.31 0.99 23.03 2.77 2.54 108.05
3 Cheese sauce 6(0) 1.44 0.19 13.39 0.53 1.25
3 Cheese sauce (spiked) 7(1) 3.77 0.45 11.96 1.26 1.29 114.37
4 Whole egg powder 8(2) 3.89 0.44 11.40 1.23 1.24
4 Whole egg powder (spiked) 8(2) 5.16 0.46 8.96 1.29 1.01 88.91
5 Multigrain cereal 10(0) 125.53 11.97 9.54 33.52 1.74

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

5 Corn cereal 10(0) 145.32 16.19 11.14 45.33 2.08
5 Corn cereal (spiked) 10(0) 155.65 26.34 16.92 73.75 3.20 63.10
6 Infant formula (powdered) 10(0) 15.84 2.20 13.91 6.16 1.86
6 NIST SRM 1846 10(0) 32.13 3.74 11.65 10.47 1.74 118.56
6 Beans with franks (60/40, w/w) 8(0) 0.87 1.02 117.59 2.86 11.56
7 Full fat soy flour 5(0) 0.69 0.10 14.85 0.28 1.24
7 Nonfat dry milk (spiked) 8(1) 2.08 0.28 13.35 0.78 1.32 106.06
8 Cottage cheese (spiked) 10(0) 7.49 1.48 19.76 4.14 2.37 103.22
9 Canned tuna in oil 7(1) 1.70 0.20 11.60 0.56 1.11
9 Canned tuna in oil (spiked) 9(1) 3.68 0.33 9.01 0.92 0.97 97.84
Avg. 100.35 ± 16.43

a
Number of laboratories retained after outliers removed. Number of outlier laboratories shown in parentheses.

C. Reagents (b) Vitamin E acetate.—Aldrich Chemical Co. (Milwau-

kee, WI).
(a) (1) Certified vitamin A acetate concentrate (c) Mixed tocopherol standard.—Prepare 50 + 50 chro-
(USP).—Equivalent to ca 30 mg of retinol/g of oil (content matographic resolution test mixture of alpha- and beta- or al-
certified by United States Pharmacopeia, Rockville, MD pha- and gamma-tocopherol in the mobile phase, at concentra-
20852 USA; www.usp.org); or (2) Retinyl palmitate, tion level of highest alpha-tocopherol standard.
all-trans.—Fluka Chemical Co., Ronkonkoma, NY, (d) Acetic acid.—Glacial.
+1-800-358-5287. Request Certificate of Lot Analysis when (e) Methanol.—HPLC grade.
ordering. If manufacturer’s certification is unavailable, or pu- (f) Ethanol.—95%.
rity of standards needs to be verified, test vitamin A palmitate (g) Tetrahydrofuran.
purity as follows: Dissolve 50 mg (record to the nearest (h) Hexane.
0.1 mg) of retinol palmitate standard in 2-propanol (i) Pyrogallic acid.—Crystals.
(UV-spectroscopy grade) in a 500 mL flask and dilute to vol- (j) Vitamin A mobile phase.—Combine 860 mL methanol
ume. Dilute 10 mL of this solution to 100 mL with 2-propanol (HPLC grade) and 140 mL water. Mix well. Stir overnight to
(final concentration is ca 10 mg/L). Measure maximum degas or mechanically degas prior to use.
absorbance obtained at 325–328 nm using 1 cm pathlength (k) Vitamin E mobile phase.—Combine 940 mL methanol
cell and 2-propanol as a blank. Calculate the purity of the (HPLC grade) and 60 mL water. Mix well. Stir overnight to
retinol palmitate as follows (6): degas or mechanically degas before use.
(l) THF–ethanol (50 + 50).—Combine 500 mL
Percent purity = (ABS × 5 × 106) / (960 × W) tetrahydrofuran and 500 mL 95% ethanol. Mix well.
(m) Potassium hydroxide solution, 50%.—Slowly add
where ABS = absorbance maximum; 960 = absorbance of 500 g of KOH pellets to 500 mL water contained in a 2 L thick
pure retinol palmitate (1% solution in 1 cm cell); W = weight wall Erlenmeyer flask. (Caution: The solution gives off sub-
of test portion in mg; and 5 × 106 = combined dilution factors, stantial heat while KOH is dissolving; add the KOH in 100 g
conversion to 1% equivalent solution, and conversion to %. portions while the flask is being cooled with cold water. Swirl
Store retinol palmitate standard at 0–4°C to allow for easier the flask gently to aid in dissolution of the KOH. Store in glass
handling while weighing. container with cork stopper.)
428 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002

Table 2001.13D. Intralaboratory study results for determination of vitamin E in foods by liquid chromatography
Mean,
Youden pairs Labsa mg/100 g sr RSDr, % sR RSDR, % r R HORRAT

Corn cereal (spiked) and corn cereal 10(0) 150.45 12.03 8.00 21.86 14.53 33.68 61.21 2.73
Multigrain cereal and NIST SRM 1846 10(0) 79.24 8.97 11.32 8.97 11.32 25.12 25.12 1.93
Infant formula (powdered) and 7(1) 14.76 0.22 1.50 2.24 15.16 0.62 6.27 2.01
margarine/butter (50/50 mix; spiked)
Cottage cheese (spiked) and whole egg 10(0) 6.40 0.51 7.94 0.65 10.17 1.43 1.82 1.19
powder (spiked)
Margarine/butter (50/50 mix) and 10(0) 3.67 0.44 12.12 0.53 14.35 1.23 1.48 1.54
canned tuna in oil (spiked)
Whole egg powder and chicken gravy 8(0) 4.10 0.71 17.33 0.77 18.74 1.99 2.16 2.05

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

(canned; spiked)
Cheese sauce (spiked) and nonfat dry 7(0) 2.90 0.21 7.36 .36 12.23 0.59 1.01 1.27
milk (spiked)
Canned tuna in oil and cheese sauce 6(0) 1.571 0.033 2.10 0.201 13.04 0.09 0.56 1.23

a
Number of laboratories retained after outliers removed. Number of outlier laboratories shown in parentheses.

(n) Vitamin A working standard (ca second 125 mL low actinic Erlenmeyer flask. Add 33 mL of
15 mg/mL).—(1) Using USP standard.—Weigh 50 mg vita- 95% ethanol. Proceed to addition of pyrogallic acid.
min A acetate concentrate into a 100 mL low actinic volumet- Prepare low standard by pipeting 0.5 mL vitamin A work-
ric flask. Record weight to nearest 0.1 mg. Record concentra- ing standard and 2 mL vitamin E working standard into a third
tion in mg/g per USP certification. Add small amount of 125 mL low actinic Erlenmeyer flask. Add 37.5 mL of 95%
acetone (<3 mL) to aid dissolution. Dilute to volume with ethanol. Proceed to addition of pyrogallic acid.
95% ethanol. Store at 4°C in dark. Solution is stable for
Grind solids to pass a 40 mesh sieve. Blend liquid or wet
2 weeks. (2) Using retinyl palmitate.—Weigh 55 mg of
materials to homogeneity and store ≤4°C in the dark.
retinyl palmitate into a 100 mL low actinic volumetric flask.
Record weight to nearest 0.1 mg. Record purity per supplier To prepare low fat (<40% fat) test samples, weigh enough
certification or purity test. Add pea-sized piece of pyrogallic test sample (≤5 g) to give ca 50 µg of vitamin A and/or 1.0 mg
acid, ca 50 mg. Dissolve and dilute to volume with hexane. vitamin E into a 125 mL low actinic Erlenmeyer flask. For test
Pipet 5 mL of solution to second 100 mL low actinic flask and samples high in sugar, add 3 mL water and disperse the test
dilute to volume with 95% alcohol. Store at 4°C in dark. Solu- portion as a slurry. Add 40 mL of 95% ethanol.
tion is stable for 2 weeks. To prepare high fat test samples, weigh test sample (≤2 g) to
(o) Vitamin E stock solution.—ca 500 µg/mL. Weigh give ca 50 µg of vitamin A and/or 1.0 mg vitamin E into a 125 mL
50 mg vitamin E acetate into 100 mL low actinic volumetric low actinic Erlenmeyer flask. Add 40 mL of 95% ethanol.
flask. Record weight to nearest 0.1 mg. Add small amount of Add a pea-sized piece (ca 50 mg) of pyrogallic acid (anti-
acetone (<3 mL) to aid dissolution. Dilute to volume with oxidant) to each standard and test flask. Add a glass bead or
95% ethanol. Store at 4°C in dark. Solution is stable for boiling stone to promote even boiling.
1 month. Swirl all flasks to ensure that all materials are thoroughly
(p) Vitamin E working standard.—ca 50 µg/mL. Pipet dispersed in the solution.
10 mL vitamin E stock solution, (o), into 100 mL low actinic
Turn on N flow and ensure N atmosphere for all flasks be-
volumetric flask. Dilute to volume with 95% ethanol. Store at
fore and while refluxing.
4°C in dark. Solution is stable for 1 month.
Pipet 10 mL of 50% KOH solution into each flask and im-
D. Extraction and Saponification mediately place flask on hot plate under reflux condenser.
Swirl.
Turn on hot plate to preheat. Start and adjust cooling water
flow to precool reflux condensers. Reflux 45 min. Swirl flasks every 10 min.
Prepare high standard by pipeting 5 mL vitamin A working Remove reflux flasks from hot plate, stopper with corks, and
standard, C(n), and 10 mL vitamin E working standard, C(p), quickly cool flasks to room temperature using cold water or ice
into a 125 mL low actinic Erlenmeyer flask. Add 25 mL of water.
95% ethanol. Proceed to addition of pyrogallic acid. Pipet 10 mL of glacial acetic acid into each flask to neutral-
Prepare intermediate standard by pipeting 2 mL vitamin A ize the KOH. Mix well and let flasks cool again to room tem-
working solution and 5 mL vitamin E working solution into a perature.
 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002 429

Quantitatively transfer the solution in each flask to a cedure; 0.5458 = ratio of retinol to retinyl palmitate molecular
100 mL low actinic volumetric flask using THF–95% ethanol weights; and 200 = combined dilution factors/ conversion
(50 + 50). Dilute to volume with the same solvent mixture. from mg to µg.
Stopper and invert volumetric flask 10 times. The RFA values of the low, medium, and high standards
Allow flasks to set for at least 1 h at room temperature and should agree with each other within 3% relative since the de-
preferably overnight in refrigerator to precipitate fatty acid tector response should be linear across this concentration
salts formed during saponification. In some cases, range. Use an average of RFA values calculated from high,
centrifugation may reduce settling time. medium, and low standards for test sample quantitation.
Measure the peak heights or areas corresponding to retinol
E. Determination
(vitamin A) in the test sample extracts. The 13-cis isomer of
Start HPLC system(s) and allow to warm up and retinol (eluting immediately proceeding the all trans isomer)
equilibrate for a minimum of 30 min with mobile phase flow- might be present in some test samples. Measure the 13-cis
ing at flow rate of 1.0 mL/min. peak also.
Inject vitamin A standards that have been taken through Multiply the height or area of the 13-cis retinol peak by 1.08

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

saponification onto HPLC system. Adjust mobile phase to (to compensate for difference in absorbance compared to the
achieve a resolution of 1.5 or better for cis and trans forms. All trans isomer).
trans retinol should elute in ca 9 min. Inject chomatographic Add the corrected peak height or area for the 13-cis isomer
resolution test mixture onto vitamin E LC system. Adjust mo- to that of the all-trans isomer to give total test sample peak
bile phase to achieve resolution of 1.5 or better for the 2 forms height or area. Calculate the concentration of vitamin A (in
of tocopherol present in the resolution test mix. µg/g as retinol) using the following equation:
alpha-Tocopherol should elute in ca 10 min.
Inject high, medium, and low standards. Adjust detector RFA × PkHTSPLE × 100
sensitivity to give peak heights of 50–90% of full scale for the Vitamin A, µg/g (as retinol) =
W
vitamin of interest at the high standard. Repeat injection of
standard until peak height(s) are reproducible.
where RFA = response factor for vitamin A; PkHTSPLE = total
Inject test solutions. Intersperse with standard solution in-
test sample peak height or area of all trans and 13-cis retinol;
jections after every 9 tests. [If retinol or alpha-tocopherol in
100 = dilution volume of test portion, mL; and W = weight of
test exceeds the peak height of their respective high standard
test portion, g.
by more than 25%, dilute test solutions using a solution of
10 mL 50% KOH, 40 mL of 95% ethanol, 10 mL glacial acetic (b) Vitamin E.—Calculate mg/g vitamin E (as al-
acid, and 40 mL THF–95% ethanol (50 + 50).] pha-tocopherol acetate) as follows: Measure peak height or
area of alpha-tocopherol in the standards. Determine response
F. Calculations factor for vitamin E (RFE) using the following equation:
(a) Vitamin A.—Calculate µg/g of vitamin A (as retinol)
mg std × mL std
as follows: Measure the peak heights or areas of the standards. RFE =
(1) Using USP standard.—Determine the response factor PkHTstd × 100000
for vitamin A (RFA) using the following calculation:
where PkHTstd = peak height or area of standard from
mg std × mL std × conc std chromatogram; mLstd = mL of working standard used in pro-
RFA =
PkHTstd × 10000 cedure; mgstd = mg alpha-tocopherol acetate standard
weighed; and 100 000 = combined dilution factors in vitamin
where PkHTstd = peak height or area of standard from E standard.
chromatogram; mLstd = mL of working standard used in pro- The RFE values of the low, medium, and high standards
cedure; concstd = concentration of USP vitamin A (as retinol) should agree with each other within 3% relative since the de-
per USP certification (mg/g); mgstd = mg of USP standard tector response should be linear across this concentration
weighed in reagents section; 10000 = combined dilution fac- range. An average of the low, medium, and high RFE stan-
tors for vitamin A standard. dards should be used for test sample quantitation.
(2) Using retinyl palmitate.—Determine the response fac- Measure peak height or area of peak corresponding to al-
tor for vitamin A (RFA) using the following calculation: pha-tocopherol on chromatogram of test sample extract.
Use the following equation to calculate the concentration
mg std × mL std × purity std × 0.5458
RFA = of vitamin E (as alpha-tocopherol) in mg/g:
PkHTstd × 200
Vitamin E, mg/g (as alpha-tocopherol acetate) =
where puritystd = percent purity certified by supplier or deter-
mined, divided by 100; mgstd = mg of retinyl palmitate
weighed; PkHTstd = peak height or area of standard from RFE × PkHTSPLE × 100
chromatogram; mLstd = mL of working standard used in pro- W
 430 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002
Table 2. Collaborative study data for vitamin A
Laboratory

Spike level,
Food matrix sector Vitamin A samples µg/100 g 1 2 3 4 5 6 7 8 9 10 11 12

1 Margarine/butter (50/50 mix) 874 806 732 871 860 960 820 934 832 948 781 867

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

1 Margarine/butter (50/50 mix; spiked) 563 1470 1283 969a 1480 1480 1570 1390 1492 1347 1530 4278a 1530
2 Chicken gravy (canned) 10
2 Chicken gravy (canned; spiked) 135 140 92 127 158 140 140.2 113 141 342a 129
3 Cheese sauce 39 64 80 29.5 35
3 Cheese sauce (spiked) 214 270 258 304 296 300 250 222 227 263 338 258
4 Whole egg powder 176 67 144 188 160 186.8 118 191 293 96
4 Whole egg powder (spiked) 288 467 155 680 433 430 180 337.4 479 881 506 282 283
5 Corn cereal 968 1008 766 970 950 100a 847.7 894 1043 1063 910 981
5 Corn cereal (spiked) 415 1430 1369 1316 1340 1390 1380 1279 1455 1465 1551 359a 1380
5 Multigrain cereal 599 571 500 717 620 630 627.6 841 727 778 561 441
6 Baked beans with franks (60/40 w/w) 2 395 40
6 Infant formula (powdered) 620 575 604 608 350 660 593.8 635 563 737 475 589
6 NIST SRM 1846 584 499 453 362 457 450 450 468.3 400 480 547 508 497
7 Nonfat dry milk 896 822 658 725 750 890 780.2 691 914 1005 794 882
7 Nonfat dry milk (spiked) 710 1720 1738 1924 1750 1520 1800 1625 1748 1825 1994 1245 1610
7 Full fat soy flour 15 40 363.8 208
8 Cottage cheese 36 39 84a 57.7 50 50 38.19 55 42
a a
8 Cottage cheese (spiked) 369 407 404 459 354 270 1430 392.3 491 391 512 838 430
9 Canned tuna in oil
9 Canned tuna in oil (spiked) 293 287 163 227 263 530a 260 284 266 338 351 187 260

a
Dixon outlier.
Table 3. Collaborative study data for vitamin E
Laboratory
Food
matrix sector Vitamin E samples Spike level, mg/100 g 1 2 3 4 5 6 7 8 9 10

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

1 213 Margarine/butter (50/50 mix) 4.04 2.5 3.7 4.8 —a —a 3.71 3.6 3.9
a
1 241 Margarine/butter (50/50 mix; spiked) 9.25 12.5 11.8 11.7 17.2 15 — —a 12.36 16 13.6
2 215 Chicken gravy (canned) 0.1 3.9 0.26
2 243 Chicken gravy (canned; spiked) 3.99 3.47 3.6 3.5 5.5 3.71 3.91 4.8 6
a a
3 217 Cheese sauce 1.33 1.3 1.3 — — 1.49 1.8 1.4
3 245 Cheese sauce (spiked) 1.96 3.42 3.5 3.3 6b 4.3 —a —a 3.48 4.4 3.97
b b
4 219 Whole egg powder 3.51 3.9 3.4 9 4.4 4.62 6.1 3.62 4.1 3.6
4 247 Whole egg powder (spiked) 1.42 5.32 5.2 4.9 13.7b 4.5 4.89 8.65b 5.23 6.1 5.11
5 221 Multigrain cereal 122 127 125.9 121.2 144 120.4 113.84 106 144 131

DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002 431
5 223 Corn cereal 149 154.4 147.1 146.9 134 148.3 119.54 123 174 157
5 249 Corn cereal (spiked) 16.36 163 158.5 170.4 156.2 110 152.6 153.77 116 198 178
6 225 Infant formula (powdered) 15.2 14 14.1 20 13.4 15.3 18.21 14.4 18.2 15.6
6 227 NIST SRM 1846 27 33.7 27.3 31.6 30.2 26.4 32.48 32.91 31.9 39.3 35.5
6 229 Baked beans with franks (60/40 w/w) 0.2 0.147 2.9 0.65 2.01 0.33 0.4 0.3
7 231 Full fat soy flour 0.7 0.6 0.7 —a —a 0.85 0.6
7 235 Nonfat dry milk 0.009 0.62
7 251 Nonfat dry milk (spiked) 1.96 2.01 2.3 2 1.7 2.17 5.86b 1.86 2.6 1.99
8 237 Cottage cheese 0.1 0.23 2.2
8 253 Cottage cheese (spiked) 7.26 7.25 7.2 7 4.2 8.2 6.55 9.56 7.67 8.9 8.41
9 239 Canned tuna in oil 1.61 1.6 1.5 5.6b 1.7 1.8 2.1 1.62
9 257 Canned tuna in oil (spiked) 1.92 3.55 3.1 3.4 7.4b 3.5 3.88 4.19 3.78 4 3.75

a
Laboratory failed to obtain required chromatographic resolution.
b
Dixon outlier.
432 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002

where RFE = response factor for vitamin E; PkHTSPLE = test achieved. It does not appear to be an issue with the protein
sample peak height or area of alpha-tocopherol; 100 = dilution content of the test sample, since test samples in adjacent sec-
volume of test portion, mL; and W = weight of test portion, g. tors 7 and 8, which have similar protein content, were ana-
Alternatively, a 3 level calibration using a zero order poly- lyzed with satisfactory results.
nomial fit (linear) can be used to calculate vitamins A and E. The Youden pair statistics approach should be applicable
Ref.: J. AOAC Int. 85, 425–432(2002) for all pairs of test samples where the variance of the pair is ex-
pected to be similar. Since the method being studied is appli-
Results and Discussion cable to all foods, similar variance would be expected for test
samples of similar analyte level, regardless of their matrix.
Vitamin A
Therefore, the Youden pair statistical approach was applied to
Collaborative study data for vitamin A were received from the data to determine the within-laboratory variability for the
13 laboratories. The data from one of the laboratories showed method. Youden pairs were set up for those test samples hav-
a systematically high bias compared to the other laboratories. ing the closest analyte levels. The results are shown in Ta-
A detailed investigation of the procedure used by that labora- ble 2001.13B. The 14 test samples can be combined into

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

tory showed that the method specified was not followed with 7 pairs. The between-laboratory variability calculated using
regard to standard preparation. The results of that laboratory the Youden pair approach is very similar to the be-
are not included in this report. Table 2 provides a tabulation of tween-laboratory variability arrived at by treating each test
the results obtained from the remaining 12 laboratories. For 4 sample singularly. As expected, the within-laboratory vari-
of the test samples, chicken gravy, cheese sauce, baked beans ability for each pair of test samples is less than the be-
with franks, and canned tuna, insufficient data were received tween-laboratory variability.
from the reporting laboratories to conduct an adequate statisti- To assess the recovery by the method, 8 test samples were
cal evaluation. Full fat soy flour, as a plant-based material, spiked with a range of vitamin concentrations. In addition,
will not have a measurable quantity of retinol (test samples NIST SRM 1846 was included in the test sample set. As a re-
were selected based on published theoretical data for both vi- sult, a recovery sample was included for each of the sectors of
tamins A and E content). All of the remaining food samples the food analysis triangle. The results of the recovery calcula-
were expected to have nutritionally significant quantities of tions are shown in Table 2001.13A. The overall recovery for
vitamin A based on published food table data. The typical LC the method was 100 ± 13%, which is very much in line with
operating limit of detection/quantitation for retinol is the expected variability of recovery for a method with an ex-
15 µg/100 g in most laboratories. It is possible that the actual pected RSDR of approximately 13% (see the Horwitz ratio
values for the 4 samples listed are <15 µg/100 g, or that the de- values in Tables 2001.13A and B).
tectors used by some of the laboratories were not adequately SRM 1846, a powdered, milk-based infant formula avail-
tuned to a sufficiently low signal-to-noise ratio to detect the able from NIST, was included as a test sample with the un-
low levels of retinol actually present. In the case of each of knowns sent to participating laboratories as a check on accu-
these 5 test samples, chicken gravy, cheese sauce, baked beans racy of the method. The NIST Certificate of Analysis lists the
with franks, canned tuna, and soy flour, at least one additional noncertified vitamin A content of the NIST sample as 5.84 ±
test sample from that food sector was present for which valid 0.68 mg/kg (584 ± 68 µg/100 g), 95% uncertainty range. The
data were obtained. collaborative study results were 464 ± 31 µg/100 g, 95% un-
The AOAC guidebook for Study Directors does not pro- certainty range. This represents a significantly lower result,
vide guidance for statistical processing of individual test sam- i.e., recovery of 79.5%. Packets of SRM 1846 had been pro-
ples. It provides guidance only on statistical evaluation appli- cured at least 6 months in advance of the study and stored, un-
cable to duplicates and/or Youden pairs. Therefore, we opened, in a dark cabinet in an office where the temperature
applied the Dixon test to each of the individual test samples typically ranged from 20 to 25°C (68 to 77°F) per storage in-
having a sufficient number of data points to obtain the be- structions supplied with the SRM. Because the level of
tween-laboratory statistical data for the test samples. The re- cis-retinol was somewhat elevated beyond what might be ex-
sults are shown in Table 2001.13A. Dixon outliers are noted in pected in the SRM, the authors decided to investigate the pos-
Table 2. The results obtained were good across all sectors ex- sible loss of retinol on storage of this material. New packets of
cept sector 4 (33–67% fat, 33–67% protein, and 0–33% carbo- SRM 1846 were purchased and analyzed side by side with re-
hydrates) represented by whole egg samples. The variability tained packets of SRM that had now been stored for an addi-
of results [expressed as relative standard deviations (RSDs) at tional 8 months (at least 14 months total). The NIST samples
various concentrations] is in agreement with what one would were analyzed at ca 2 weeks, 5 weeks, and 6 months after re-
expect as indicated by the Horwitz ratio (HORRAT column) ceipt of the new packets. The results for the stored NIST sam-
for expected collaborative results relative to level of analyte. ple were 80.7, 78.6, and 81.1% compared to the newly re-
(Normally one expects a HORRAT of <2.) For whole egg ceived materials, respectively (average 80.1%). Average
samples, it is uncertain whether the particular fat, carbohy- concentration for stored NIST samples was 465 µg/100 g. The
drate, or protein ratio present is the source of the analytical average concentration was 580 µg/100 g for the newly pur-
variability, if egg as a matrix is difficult to digest and extract, chased NIST samples. The actual cause of the reduced level of
or if homogeneity of the whole egg powder sample was not vitamin A in the SRM used in the collaborative study is un-
 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002 433

known, but the level found is consistent with the comparative beans with franks had an exceptionally high variability. An in-
result in the subsequent study. Although the office where the tuitive review of the data for baked beans and franks would
SRM was stored prior to the collaborative study was cli- cause one to reject the data from laboratories 4 and 7; how-
mate-controlled, the area of the building immediately adjacent ever, the results of the laboratories were not removed by the
to the office is a food production pilot plant facility. Investiga- statistical outlier test.
tion with the sanitation officer revealed that this food produc- Again, as with vitamin A, the Youden pair statistical ap-
tion area was periodically heated well over 100°F on weekends proach was applied to the data to determine the
for pest control purposes. It may be that the cabinet in the office, within-laboratory variability for the method. Youden pairs
against the wall adjacent to the food production facility, experi- were set up for those test samples having the closest analyte
enced periods of high temperature during the time the NIST levels. The results are shown in Table 2001.13D. The 16 sam-
sample was stored there, thus exposing the NIST sample used ples can be combined into 8 pairs. The between-laboratory
in the collaborative study to degradation temperatures. variability calculated using the Youden pair approach is very
similar to the between-laboratory variability arrived at by
Vitamin E treating each sample singularly. As expected, the

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

Collaborative study data for vitamin E was received from within-laboratory variability for each pair of test samples is
10 laboratories. Table 3 provides a tabulation of results ob- less than the between laboratory variability.
tained from all 10 laboratories. For 5 of the test samples, mar- To assess the recovery by the method, 8 test samples were
garine/butter mixture, spiked margarine/butter mixture, spiked with a range of vitamin concentrations. In addition,
cheese sauce, spiked cheese sauce, and full fat soy flour, re- NIST SRM 1846 was included in the test sample set. As a re-
sults from several laboratories were at odds with results from sult, a recovery sample was included for each of the sectors of
the remainder of the laboratories. Review of the the food analytical triangle. The results of the recovery calcu-
chromatograms and communication with participating labora- lations are shown in Table 2001.13C. The overall recovery for
tories indicated that, although specified in the method, the rec- the method was 100 ± 16%. This is very much in line with the
ommended adequate resolution of the various tocopherols was expected variability of the recovery for a method that has an
not achieved before analyses were conducted, i.e., several lab- expected RSDR of ca 9% (see the Horwitz ratio values in Ta-
oratories failed to ascertain adequate chromatographic resolu- bles 2001.13C and D).
tion. Excellent agreement was achieved between laboratories Collaborators’ Comments
that routinely analyze for mixed tocopherols. Although the
method submitted to the collaborators specified a minimum Vitamin A.—Collaborating laboratories reported no diffi-
resolution for separation of tocopherol isomers, some collabo- culties regarding the vitamin A assay. In general, collabora-
rators misinterpreted the instructions; therefore, the explana- tors found that the method performed well, and they appreci-
tion of the means of establishing adequate resolution was ex- ated the easy preparation procedure. One collaborator
panded by adding more detail to the method. The data from disagreed with some of the terminology used in the calcula-
these particular laboratory/sample combinations, therefore, tions, but did not offer alternative suggestions.
had to be eliminated (see Table 3, footnote a). As a result, for 2 Vitamin E.—Conditions proposed for vitamin E collabora-
of the samples, cheese sauce and full fat soy flour, data for the tive study are the same as routinely used in the laboratory of
statistical evaluation are limited. The number of samples is one of the collaborators. The detector is linear well beyond
noted in Table 2001.13C. For the nonfat dry milk sample and calibrated range. The procedure was easy to follow and does
the cottage cheese sample, only 2 and 3 laboratories reported not involve lengthy preparation times.
detecting alpha-tocopherol, respectively. The typical LC op-
erating limit of detection/quantitation for alpha-tocopherol is Recommendations
0.1–0.2 mg/100 g in most laboratories. It is most likely that the
actual values for these test samples are <0.2 mg/100 g. For Collaborators who followed the instructions and method
these 2 test samples, at least one additional test sample from provided were successful in performing the procedure, and
that food sector was present for which valid data were ob- there is a need for an LC method for vitamins A and E across
tained. All of the remaining food samples were expected to all food categories. As a result, the Study Coordinators recom-
have measurable quantities of vitamin E based on published mend that this method for the determination of vitamins A and
food table data. E in foods by liquid chromatography be adopted Official First
As with vitamin A, we applied the Dixon test to each test Action by AOAC INTERNATIONAL.
sample having a sufficient number of data points to obtain the
between-laboratory statistical data for the samples. The results Acknowledgments
are shown in Table 2001.13C. Dixon outliers are noted in Ta-
ble 3. The variability of results (expressed as RSDs at various We thank Paul Wehling (Medallion Laboratories, General
concentrations) are generally in agreement with what one Mills, Inc., Minneapolis, MN) for assistance with statistical
would expect as indicated by the Horwitz ratio (HORRAT evaluations and the following collaborators:
column) for expected collaborative results relative to level of Sami Al-Hasani, ConAgra Frozen Foods Co., Columbia,
analyte. (Normally one expects a HORRAT of <2.) Baked MO
434 DEVRIES & SILVERA: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 2, 2002

Jocelyn Alfiere, Diversified Research Labs, Inc., Donald Willis, Ralston Analytical Laboratories, St. Louis,
Markham, ON, Canada MO
Carrie Berge and Ellen Hanson, Novartis Nutrition, Min- Becky Wittrig, TPC Labs, St. Paul, MN
neapolis, MN
Claudia Boerner and Ed Vinski, Microbac Laboratories, References
Warrendale, PA
Meena Chettiar and Ross Mazal, Land O’ Lakes, Inc., St. (1) Official Methods of Analysis (2000) 17th Ed., AOAC IN-
Paul, MN TERNATIONAL, Gaithersburg, MD
Sarita Cardozo, Anresco, Inc., San Francisco, CA (2) Egberg, D.C., Heroff, J.C., & Potter, R.H. (1977) J. Agric.
Kerri Gustafson, Medallion Laboratories, General Mills, Food Chem. 25, 1127–1132
Inc., Minneapolis, MN (3) DeVries, J.W., Egberg, D.C., & Heroff, J.C. (1979) Liquid
Assad Kazeminy, Irvine Analytical Laboratories, Irvine, Chromatographic Analysis of Food and Beverages, G.
Charalambous (Ed.), Academic Press, New York, NY,
CA
Vol. 2, pp 477–497
Dana Krueger and Boro Mioc, Krueger Food Laboratories,

Downloaded from https://academic.oup.com/jaoac/article/85/2/424/5656624 by guest on 28 June 2024

(4) DeVries, J.W. (1985) in Methods of Vitamin Assay, 4th Ed.,
Cambridge, MA
J. Augustin, B. Klein, D. Becker, & P. Venugopal (Eds), John
Peter Meland and Kari Dupont, Ingman Laboratories, Inc., Wiley and Sons, New York, NY, pp 65–94
Minneapolis, MN (5) Ikins, W., DeVries, J., Wolf, W., Oles, P., Carpenter, D.,
Lisa L. Oehrl, Southern Testing Laboratory, Wilson, NC Fraley, N., & Ngsh-Ngwainbi, J. (1993) The Referee 17, 1,
6–7, AOAC INTERNATIONAL, Gaithersburg, MD
(6) Olson, J.A. (1990) Handbook of Vitamins, L.J. Machlin and
Marcel Dekker, New York, NY, pp 1–58