AGRICULTURAL MATERIALS

Crude Fat, Hexanes Extraction, in Feed, Cereal Grain, and Forage

(Randall/Soxtec/Submersion Method): Collaborative Study
NANCY J. THIEX
South Dakota State University, Oscar E. Olson Biochemistry Laboratories, Box 2170, ASC 151, Brookings, SD 57007
SHIRLEY ANDERSON
Foss North America, 7682 Executive Dr, Eden Prairie, MN 55344
BRYAN GILDEMEISTER
South Dakota State University, Oscar E. Olson Biochemistry Laboratories, Box 2170, ASC 151, Brookings, SD 57007
Collaborators: W. Adcock; J. Boedigheimer; E. Bogren; R. Coffin; K. Conway; A. DeBaker; E. Frankenius; M. Gramse;
P. Hogan; T. Knese; J. MacDonald; J. Mller; R. Royle; M. Russell; F. Shafiee; B. Shreve; J. Sieh; M. Spann; E. Tpler;
M. Watts
A method for determining crude fat in animal feed,
cereal grain, and forage (plant tissue) was col-
laboratively studied. Crude fat was extracted from
the animal feed, cereal grain, or forage material
with hexanes by the Randall method, also called
the Soxtec method or the submersion method. The
use of hexanes provides for an alternative to di-
ethyl ether for fat extractions. The proposed sub-
mersion method considerably decreases the ex-
traction time required to complete a batch of
samples compared to Soxhlet. The increase in
throughput is very desirable in the quest for faster
turnaround times and the greater efficiency in the
use of labor. In addition, this method provides for
reclamation of the solvent as a step of the method.
The submersion method for fat extraction was pre-
viously studied for meat and meat products and
was accepted as AOAC Official Method 991.36.
Fourteen blind samples were sent to 14 collabora-
tors in the United States, Sweden, Canada, and
Germany. The within-laboratory relative standard
deviation (repeatability) ranged from 1.23 to 5.80%
for crude fat. Among-laboratory (including within)
relative standard deviation (reproducibility) ranged
from 1.88 to 14.1%. The method is recommended
for Official First Action.
T
he Randall (1) or submersion method for fat extraction
is an AOAC Official Method for meats and meat prod-
ucts (2). Its use is also widespread in feed laboratories
to determine crude fat in feed, grain, and forage. Approximately
1/3 of the laboratories reporting crude fat results on animal feed
to the Association of American Feed Control Officials
(AAFCO) Check Sample Program are reporting fat values ob-
tained using this method. It therefore seemed appropriate that
this method should be collaborated for animal feed, cereal
grain, and forage in an attempt to establish the method as an of-
ficial method and to bring the AOACOfficial Methods of Anal-
ysis current with what is practiced in todays laboratories.
While seeking collaborators for the concurrent study using
diethyl ether as a solvent (see this issue, p. 888), it became ap-
parent that few laboratories actually use diethyl ether because
of safety considerations with the potential for peroxide forma-
tion in ether. Aneed was evident for an official alternative sol-
vent to diethyl ether for laboratories performing fat extrac-
tions. Petroleum ether, which is not an official solvent for
animal feed, was found to be in wide use for fat extractions
(numerous personal communications). Therefore, compara-
bility tests for petroleum ether and other potential solvents
were performed.
Ruggedness Testing
Ruggedness tests (3) were performed as part of the method
validation process. Variables studied were predry time (2 vs
4 h); boil time (20 vs 40 min); solvent (diethyl ether vs petro-
leum ether); rinse time (30 vs 60 min); test portion weight
(1 vs 3 g); extraction cup dry time (2 vs 4 h); and solvent drip
rate (2 vs 4 drops/s). Ruggedness tests were performed on
3 feed materials in 3 laboratories (see the concurrent study in
this issue, p. 888).
Solvent Comparability
Single Laboratory
Petroleum ether, pentanes, and hexanes were tested for
comparability to diethyl ether in an effort to find an alternative
to diethyl ether for laboratories desiring to avoid use of a sol-
THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003 899
Submitted for publication May 2003.
The recommendation was approved by the Methods Committee on
Feeds, Fertilizers, and Related Agricultural Topics as First Action. See
Official Methods Program Actions, (2003) Inside Laboratory
Management, May/June issue.
Corresponding authors e-mail: nancy_thiex@sdstate.edu.
vent with the potential for peroxide formation. The solvents
chosen were petroleumether, pentanes, and hexanes. Solvents
were chosen for comparison based upon boiling point, boiling
point range, polarity, price, and (in the case of petroleum
ether) current usage. Results of intralaboratory comparability
tests using 10 feed and forage materials are shown in Table 1.
Correlation coefficients, slope, and bias for petroleum ether,
pentanes, and hexanes (compared to diethyl ether) are 0.9998,
0.9947, and 0.3179; 0.9997, 0.9965, and 0.5129; and 0.9999,
0.9852, and 0.1181, respectively. The inclusion of a 100% fat
supplement tended to mask the effect of solvent on lower fat
materials; therefore, the statistics were recalculated omitting
the high fat supplement. Correlation coefficients, slope, and
bias for petroleum ether, pentanes, and hexanes (compared to
diethyl ether and omitting the high fat supplement) were
0.9949, 0.9645, and 0.4662; 0.9938, 1.0216, and 0.3941; and
0.9970, 0.9858, and 0.1661, respectively. From these data,
hexanes appear to be the best match to diethyl ether.
Interlaboratory
Comparability tests described above were repeated using
3 laboratories and 6 feed and forage materials. Results of
interlaboratory comparability tests are shown in Table 2. Cor-
relation coefficients, slope, and bias for petroleum ether, hex-
anes, and pentanes (compared to diethyl ether) are 0.9998,
1.0069, and 0.3168; 0.9999, 1.0139, and 0.1449; and
0.9997, 1.0274, and 0.3236, respectively.
Once again, the inclusion of a 100%fat supplement tended
to mask the effect of solvent on lower fat materials; therefore
the data are repeated in Table 2, omitting the high fat supple-
ment. Correlation coefficients, slope, and bias for petroleum
ether, hexanes, and pentanes (compared to diethyl ether and
omitting the high fat supplement) are 0.9878, 1.0077, and
0.3193; 0.9925, 1.0201, and 0.1672; and 0.9880, 1.0274,
and 0.4141, respectively.
Based upon these results and discussions of these results at
AAFCO Laboratory Methods and Services Committee Meet-
ings, it was decided that the most suitable alternative to diethyl
ether is hexanes. Petroleum ether is already in common use as
an alternative, but it shows a low bias for some materials, in-
cluding forages. There are also problems with consistency of
the solvent from manufacturer to manufacturer, lot-to-lot, and
country-to-country, as a fat solvent. Some of this may be due
to the wide range of components of the solvent. The wide boil-
ing point range also makes it a poor choice for recovery and
reuse of the solvent.
900 THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003
Table 1. Comparability of 3 solvents to diethyl ether (intralaboratory)
Sample description Diethyl ether Petroleum ether Pentanes Hexanes
Crude fat, %
Fat supplement 99.02 99.13 98.94 99.35
Alfalfa hay 1.38 1.06 1.01 1.51
Soybean meal 1.17 1.15 1.00 1.32
Corn silage 2.04 1.94 1.85 2.02
Beet pulp 0.35 0.41 0.23 0.41
Pig starter 8.63 8.42 8.40 8.36
Pig starter, ww
a
8.09 7.65 7.15 7.62
Meat and bone meal 10.72 10.39 10.73 11.02
Extruded supplement 18.83 19.34 17.73 19.06
Cattle protein supplement, 7% urea 3.64 3.20 3.21 3.35
Cattle protein supplement, 7% urea, ww
a
2.57 2.10 1.98 2.49
Corn 3.31 1.99 1.92 2.60
Summary statistics omitting fat supplement
Correlation R 0.99746 0.99690 0.99851
R
2
0.99493 0.99381 0.99702
Slope 0.9645 1.0216 0.9858
y-Intercept 0.4662 0.3941 0.1661
Std. error 0.42 0.46 0.32
Sum 60.73 57.65 55.21 59.75
Average 5.52 5.24 5.02 5.43
a
ww = Water-wash.
Collaborative Study
The Study Directors and 11 collaborating laboratories con-
ducted the collaborative study. Laboratories represented a va-
riety of types that would routinely use the proposed method,
including research, commercial, industrial, and state regula-
tory laboratories. Samples were sent to 4 laboratories outside
the United States, and results on study samples were received
from 3 of them. Participants received no compensation. Fa-
miliarization samples were sent to each collaborator to be ana-
lyzed before the test samples to acquaint them with the
method and to ensure that the laboratory was capable of han-
dling the test samples.
Collaborating laboratories were asked to analyze 14 animal
feed, cereal grains, and forage materials as blind duplicate
pairs, resulting in 28 test samples. Ablank material (cellulose)
was also included as a blind duplicate pair and labeled as a
sample.
Study materials were chosen to be representative of differ-
ent feed, cereal grain, and forage materials (Table 3). All sam-
ples were natural or real world; none were spiked. Samples
were coded at random with no preselection from order of pre-
sentation. Approximately 20 g of each material was provided,
which was in excess of the amount needed to complete the
study. Two materials (a urea-containing feed and a high-sugar
feed) were identified as requiring a water prewash. Partici-
pants were informed which samples were low or high enough
in fat concentration to require weighing a test portion larger or
smaller than 2 g. Concentration ranges for the sample ranged
from a blank (<0.5% fat) to nearly 100% fat.
Study materials were prepared as follows: The meat
meal/hulls mixture, calf feed medicated, broiler starter, calf
starter medicated, dehydrated alfalfa, medicated goat feed, and
swine feed were donated by the AAFCO Check Sample Pro-
gram. These materials were used without further grinding. The
corn silage was ground in a Foss Tecator Cyclotec Mill (Foss
THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003 901
Table 2. Comparability of 3 solvents to diethyl ether (interlaboratory)
Sample type Lab No. Diethyl ether Petroleum ether Hexanes Pentane
Crude fat, %
Fat supplement 1 97.64 99.23 99.45 99.11
2 97.10 97.33 97.91 96.53
3 97.51 96.75 98.49 96.27
Alfalfa hay 1 1.41 1.05 1.54 1.04
2 1.04 1.07 1.25 1.06
3 1.41 0.88 1.28 0.81
Beet pulp 1 0.35 0.41 0.38 0.30
2 0.33 0.24 0.19 0.18
3 0.23 0.06 0.18 0.09
Meat and bone meal 1 10.45 10.49 10.94 11.05
2 10.42 10.77 10.54 10.28
3 10.68 9.94 10.59 10.30
Cattle protein supplement 1 2.47 2.13 2.58 2.08
2 3.29 3.06 3.07 2.86
3 3.54 2.77 2.98 2.69
Corn 1 3.32 2.00 2.37 1.98
2 3.68 3.66 3.65 3.62
3 3.78 3.51 3.47 3.39
Summary statistics omitting fat supplement
Correlation R 0.99389 0.99623 0.99398
R
2
0.98782 0.99247 0.98799
Slope 1.0077 1.0201 1.0274
y-intercept 0.3193 0.1672 0.4141
Std. error 0.43 0.34 0.44
Sum 56.40 52.04 55.02 51.73
Average 3.76 3.47 3.67 3.45
North America, Eden Prairie, MN) with a forage head to pass
through a 1 mm screen. The birdseed, texturized feed, and beef
concentrate pellets were ground in a Retsch ZM100 Ultra Cen-
trifugal mill (Retsch, Haan, Germany) to pass through a
0.75 mm sieve. The high oil corn was ground in a Wiley Mill
(Thomas Scientific Corp., Swedesboro, NJ) through a 1.0 mm
screen. The fat and cellulose required no grinding. No further
grinding was necessary by the collaborating laboratories for
any of the study materials. All materials were split in a Fritsch
Rotary Sample Divider Model Laborette 27 (Fritsch, c/o Gilson
Co., Inc., Lewis Center, OH). They were stored in polyethylene
bags in ~20 g quantities.
Uniformity (homogeneity) of the test sample sets was veri-
fied by selecting 3 bags at random for each material and ana-
lyzing them by the proposed method in the Study Directors
laboratory. Relative standard deviations (RSDs) are shown in
Table 3. An acceptance criterion of an RSDof ca 2%was con-
sidered acceptable for the study materials.
Collaborators were asked to report results on an as is ba-
sis, to determine analysis in single for each sample, and to re-
port data to 4 significant figures. In addition, collaborators
were asked to complete a Study Survey. In addition to famil-
iarization and study samples, thimbles and cotton were pro-
vided to collaborators.
Note: The study was conducted concurrently using diethyl
ether as the extraction solvent. The concurrent study appears
as a companion article in this issue of J. AOAC Int. (see
p. 888).
AOAC Official Method 2003.06
Crude Fat in Feeds, Cereal Grains, and Forages
Randall/Soxtec/Extraction-Submersion Method
First Action 2003
[Applicable to the analysis of forages, cereal grains, and
animal feeds other than baked or expanded products, dried
milk or milk products, fishmeal, or oilseeds at concentrations
from 0.5 to 100% fat. It is applicable to the same matrixes as
AOAC Official Methods 920.39 (see 4.5.01) and 930.09 (see
3.5.07).]
Caution: Store solvents in metal containers in solvent cabi-
net or solvent room that conforms to applicable
safety legislation. Ethers and hexanes are ex-
tremely flammable. Have no open flames in the
laboratory where the analysis is being performed.
Avoid inhaling vapors. Use solvents in a properly
operating hood equipped with explosion-proof
lighting, wiring, and fan. Diethyl ether has the
potential to form shock-sensitive, explosive per-
oxides with age. Check each new container of
ether for peroxides when it is opened. Also check
partial containers of ether that have not been used
for several months before using them again. Do
not use ether that contains peroxides. Dispose as
hazardous material. Stabilized ether may be used.
Ground electrical equipment and maintain in
proper working order. Follow manufacturer rec-
ommendations for installation, operation, and
safety of all extraction equipment. Make sure all
solvent is evaporated from cups before placing
them in the oven to avoid a fire or explosion.
See Table 2003.06 for results of interlaboratory study sup-
porting acceptance of the method.
A. Principle
The Randall modification of the standard Soxhlet extrac-
tion submerges the test portion in boiling solvent, reducing the
time needed for extraction. The solvent dissolves fats, oils,
pigments, and other soluble substances, collectively termed
crude fat.
A dried, ground test portion is extracted by a 2-step pro-
cess: In the first step, the thimble containing the test portion is
immersed into the boiling solvent. The intermixing of matrix
with hot solvent ensures rapid solubilization of extractables.
The thimble is then raised above the solvent and the test por-
tion is further extracted by a continuous flow of condensed
solvent. The solvent is evaporated and recovered by conden-
sation. The resulting crude fat residue is determined
gravimetrically after drying.
The solubility characteristics of different solvents may re-
sult in slight differences in crude fat results. For this reason,
the report should reflect the solvent used. Example: % Crude
Fat, Ether Extraction; % Crude Fat, Hexanes Extraction.
902 THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003
Table 3. Categories represented by study materials
and uniformity testing results
Material
Animal
feed Forage Grain
Fat, %
RSD
Meat meal/hulls mixture 1.4
Calf feed medicated 0.80
Broiler starter 1.5
Mixed bird seed 0.86
Calf starter medicated 0.49
Corn silage 1.1
Dehydrated alfalfa 1.2
High oil corn 1.1
Medicated goat feed 0.29
Texturized feed 1.1
Feedlot concentrate pellets 1.9
Fat supplement 0.15
Swine feed 1.2
Cellulose blank (Solka-Flok)
B. Apparatus
(a) Solvent extraction system.Multiple position extrac-
tion unit conducting 2-stage Randall extraction process with
solvent recovery cycle, with Viton or Teflon

seals compati-
ble with ether or hexanes.
(b) Thimbles and stand.Cellulose thimbles and stand to
hold thimbles.
(c) Extraction cups.Aluminum or glass. (Extraction
temperature settings may differ; consult manufacturers oper-
ating instructions.)
Items (a)(c) are available as Soxtec systems from Foss
Tecator, Hgans, Box 70, SE-263 21, Sweden; Foss North
America, 7682 Executive Dr, Eden Prairie, MN 55344, Tel:
+1-952-974-9892, Fax: +1-952-974-9823, info@fossnorth
america.com; or other manufacturers of Randall-type extrac-
tion systems.
C. Reagents
(a) Hexanes.Fisher H291 (UN1208; www.fishersci.com),
or equivalent; boiling range: 4C including 68.7C.
(b) Anhydrous diethyl ether.Purified for fat extraction,
Fisher P/N E4492 labeled For Fat Extraction is also stabi-
lized, or E134-4 (UN1155), or equivalent. To prevent ether
fromabsorbing water, purchase it in small containers and keep
containers tightly closed. Petroleum ether cannot be substi-
tuted for diethyl ether because it does not dissolve all of the
plant lipid material.
(c) Cotton.Defatted. Soak medical grade cotton in di-
ethyl ether or hexanes for 24 h, agitating several times during
this period. Remove and air dry. Commercially available from
Foss Tecator, Part No. 1529-0009.
(d) Sand.Ashed (for ignition boats). EM SX0075-3, or
equivalent (CAS 14808-60-7).
(e) Celite 545.Foss Tecator 1900-0014, or equivalent.
D. Preparation of Analytical Sample
Grind laboratory samples to fineness that gives an RSD of
2.0% for 10 successive determinations.
RSD % = (SD/mean) 100
Fineness of 0.751 mm usually achieves this precision with
dry mixed feeds and other nonuniform materials.
E. Determination
Weigh 15 g test portions containing ca 100200 mg fat di-
rectly into tared cellulose thimbles, according to following
scheme:
Crude fat, % Test portion weight, g
<2 5
5 24
10 12
>20 1
Record weight to nearest 0.1 mg (S) and thimble number.
Dry thimbles containing test portions at 102 2Cfor 2 h.
If dried test portions will not be extracted immediately, store
in desiccator. Both solvent and test materials must be free of
moisture to avoid extraction of water-soluble components
such as carbohydrates, urea, lactic acid, and glycerol, which
will result in false high values.
THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003 903
Table 2003.06. Interlaboratory results for crude fat in animal feed, cereal grain, and forage, hexanes extraction
(submersion) method
Material Mean Lab
a
s
r
RSD
r
, % s
R
RSD
R
, % HORRAT
Dehydrated alfalfa 4.34 9(1) 0.14 3.21 0.16 3.75 1.17
Corn silage 1.91 9(1) 0.04 1.97 0.15 5.31 1.46
Mixed bird seed 7.15 9(1) 0.25 3.44 0.25 3.44 1.16
Texturized feed 2.91 10 0.09 3.07 0.18 6.27 1.84
Fat supplement 97.77 9(1) 1.29 1.32 1.84 1.88 0.94
Medicated goat feed 1.54 9(1) 0.03 1.94 0.13 8.48 2.26
Feedlot concentrate pellets 1.30 10 0.08 5.80 0.18 14.1 3.67
Cellulose (blank) 0.12 10 0.06 50.5 0.08 65.4 11.9
Calf starter medicated 2.58 10 0.09 3.52 0.14 5.61 1.62
Calf feed medicated 3.23 10 0.18 5.45 0.21 6.48 1.93
Meat meal/hulls mix 5.76 10 0.12 2.10 0.18 3.19 1.04
Swine feed 2.29 10 0.11 4.96 0.15 6.38 1.81
Broiler starter 5.99 10 0.17 2.83 0.22 3.61 1.18
High oil corn 7.63 9(1) 0.09 1.23 0.16 2.09 0.71
a
Number of laboratories retained after the number of laboratories in parentheses were eliminated.
An absorbent, such as diatomaceous earth (Celite or
Super-Cel), can be added to the test portion when high fat ma-
terials, which melt through the thimble during the predry step,
are present. Alternatively, defatted cotton can be added before
the predry step to absorb the melted fat. If the material melts at
102C, place a pretared extraction cup under the thimble dur-
ing the drying step to catch any melted fat that was unabsorbed
and escaped the thimble.
Place defatted (with same solvent to be used for extraction)
cotton plug on top of test portion to keep material immersed
during the boiling step and prevent any loss of test portion
from top of thimble. Prepare cotton plug large enough to hold
materials in place, yet as small as possible to minimize absorp-
tion of solvent. Adding the cotton plug before the 102
2C/2 h drying step is acceptable.
Place three or four 5 mmglass boiling beads into each cup,
and dry cups for at least 30 min at 102 2C. Transfer to des-
iccator and cool to room temperature. Weigh extraction cups
and record weight to nearest 0.1 mg (T).
Extract, following manufacturers instructions for opera-
tion of extractor. Preheat extractor and turn on condenser
cooling water. Attach thimbles containing dried test portions
to extraction columns. Put sufficient amount of solvent into
each extraction cup to cover test portion when thimbles are in
boiling position. Place cups under extraction columns and se-
cure in place. Make sure that cups are matched to their corre-
sponding thimble. Lower thimbles into solvent and boil for
20 min. Verify proper reflux rate which is critical to the com-
plete extraction of fat. This rate depends upon the equipment
and should be supplied by the manufacturer. A reflux rate of
ca 35 drops/s applies to many extraction systems.
Raise thimbles out of solvent and extract in this position for
40 min. Then distill as much solvent as possible from cups to
reclaim solvent and attain apparent dryness.
Remove extraction cups from extractor and place in oper-
ating fume hood to finish evaporating solvent at low tempera-
ture. (Note: Take care not to pick up any debris on bottom of
extraction cup while in hood. Let cups remain in hood until all
traces of solvent are gone.)
Dry extraction cups in 102 2C oven for 30 min to re-
move moisture. Excessive drying may oxidize fat and give
high results. Cool in desiccator to room temperature and
weigh to nearest 0.1 mg (F).
F. Calculations
% Crude fat, hexanes extract =
F T
S
100
-

% Crude fat, diethyl ether extract =

F T
S
100
-

where F = weight of cup + fat residue, g; T = weight of empty

cup, g; S = test portion weight, g.
Refs.: J. AOAC Int. 86, 902904(2003); 890893(2003)
Results and Discussion
Study materials were shipped the last week of September
2001 to 13 laboratories. Results were received from12 laborato-
ries (Table 4) over a period of 3 months, with the last set received
on December 28, 2001. One of the 13 laboratories could not pro-
vide data due to in-house issues. Laboratories were asked to pro-
vide the type of instrument used. Equipment used is described in
Table 5.
Early into the study, it became apparent that some collabora-
tors were following in-house methods rather than the method
supplied with the study. Also, a number of questions were re-
ceived as to the appropriateness of particular lots of solvents. At
this point, a survey was developed to document the manufactur-
ers of solvents, catalog numbers, and lot numbers; and to assure
the Study Directors that collaborators were following the
method as supplied. Results of the survey are shown in Table 6.
Laboratory 11 did not follow the method as mailed to collabo-
rators and did not have time to retest the materials. Deviations
from the method included not performing the water-wash re-
quired on 4 materials, and using different soak and rinse times
from those specified. Data from this laboratory were evaluated
using an XY plot, which confirmed that data from this labora-
tory should be removed from the study.
Collaborators Comments
In response to these comments and suggestions received
from collaborators concerning extraction of high-fat materi-
als, a better description of how to handle high-fat materials
was incorporated into the method.
A number of comments were received about the necessity
of the water-wash step. This step is specified in AOAC Offi-
cial Method 920.39 (4); and was therefore incorporated with-
out further investigation. To demonstrate the effect of poten-
tially interfering substances, recovery experiments were
performed. Urea and glucose equivalent to ~2% (~0.04 g),
~5% (~0.1 g), ~10% (~0.2 g), and ~15% (~0.3 g) were added
to 2 g test portions of ground shelled corn and extracted with-
out a water rinse. Test portions of ground corn spiked at the
15% level were also extracted after the water wash to check
whether any potential interference was removed. The data are
presented in Table 7 and graphical representations of the re-
sults of the experiment in Figures 1 and 2. Neither glucose nor
urea at levels <15% showed an interfering effect when test
portions were predried. Also, the water wash decreased recov-
ery of crude fat. It is recommended that urea and glucose not
be removed by water wash before the solvent extraction step
when hexanes are the solvent. In fact, the water wash step ap-
pears to lower the extraction efficiency or cause incomplete
recovery of crude fat, at least as compared to diethyl ether ex-
traction.
To determine the effect of the presence of water during ex-
traction on the various feed matrixes, the study materials were
extracted without drying and results were compared with
predried data collected in the collaborative study. Because the
study materials had dried out considerably since they were
prepared about 1 year before this experiment, they were
904 THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003
rehydrated by placing the weighed thimbles with test portion
into a desiccator charged with water and allowed to equilibrate
overnight at 55C. Recoveries of crude fat, hexanes extrac-
tion, ranged from 77.0 to 134% (Table 8). This is consistent
with the literature (5, 6). Water can decrease the efficiency of
the solvent in extraction, and/or allow for extraction of wa-
ter-soluble nonlipid components. The results confirm that the
predry step is critical in the extraction process. Recoveries of
fat for the urea-containing feed (feedlot concentrate pellets)
and the molasses-based, high-sugar content feed (texturized
feed) were 77 and 117%, respectively, with no predry. High
recoveries were also observed with the corn silage, medicated
goat feed, and meat meal/hulls mix. A low recovery (incom-
plete extraction) was observed with the urea-containing feed
(feedlot concentrate pellets).
A laboratory ranking by the test described by Youden and
Steiner (3) was used to assess bias among laboratories partici-
pating in the collaborative study. One laboratory on the high
and low end showed bias based on the laboratory ranking test.
However, the actual spread among values from all laboratories
was tight. The RSDof the sumof results on all materials for all
laboratories was 1.5%, and the relative difference between the
high- and low-ranking laboratory was <2%. The results were
close enough that the sumof results did not correspond with the
laboratory rank. It was decided that removing laboratories
based on the laboratory ranking test served no practical purpose
and all laboratories were kept in the statistical calculations.
Table 4 summarizes the study data and Table 2003.06 pro-
vides the statistics. The overall within-laboratory RSD
r
ranged from1.23 to 5.80%for feed, forage, and cereal materi-
als, and 50.5% for the cellulose blank. Among-laboratory
RSD
R
ranged from 1.88 to 14.1% for feed, forage, and cereal
materials, and 65.4% for the cellulose blank. The material
with the highest RSD in the collaborative study also had the
highest RSD in the homogeneity test (feedlot concentrate pel-
THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003 905
T
a
b
l
e
4
.
I
n
t
e
r
l
a
b
o
r
a
t
o
r
y
s
t
u
d
y
r
e
s
u
l
t
s
f
o
r
d
e
t
e
r
m
i
n
a
t
i
o
n
s
o
f
c
r
u
d
e
f
a
t
i
n
a
n
i
m
a
l
f
e
e
d
,
c
e
r
e
a
l
g
r
a
i
n
,
a
n
d
f
o
r
a
g
e
,
d
i
e
t
h
y
l
e
t
h
e
r
e
x
t
r
a
c
t
i
o
n
(
s
u
b
m
e
r
s
i
o
n
)
m
e
t
h
o
d
(
b
l
i
n
d
d
u
p
l
i
c
a
t
e
d
e
s
i
g
n
)
C
r
u
d
e
f
a
t
,
e
t
h
e
r
e
x
t
r
a
c
t
i
o
n
,
%
L
a
b
N
o
.
D
e
h
y
d
r
a
t
e
d
a
l
f
a
l
f
a
C
o
r
n
s
i
l
a
g
e
M
i
x
e
d
b
i
r
d
s
e
e
d
T
e
x
t
u
r
i
z
e
d
f
e
e
d
F
a
t
s
u
p
p
l
e
m
e
n
t
M
e
d
i
c
a
t
e
d
g
o
a
t
f
e
e
d
F
e
e
d
l
o
t
c
o
n
c
e
n
t
r
a
t
e
p
e
l
l
e
t
s
C
e
l
l
u
l
o
s
e
C
a
l
f
s
t
a
r
t
e
r
m
e
d
i
c
a
t
e
d
C
a
l
f
f
e
e
d
m
e
d
i
c
a
t
e
d
M
e
a
t
m
e
a
l
/
h
u
l
l
s
m
i
x
S
w
i
n
e
f
e
e
d
B
r
o
i
l
e
r
s
t
a
r
t
e
r
H
i
g
h
o
i
l
c
o
r
n
1
2
2
2
3
4
2
3
5
1
5
6
2
1
7
9
8
2
8
1
0
2
6
1
1
1
7
1
2
1
9
1
3
1
4
1
6
2
7
1
8
2
0
2
4
2
5
1
4
.
3
1
4
.
6
4
1
.
9
2
1
.
8
4
7
.
1
6
7
.
4
0
2
.
7
4
2
.
8
1
9
8
.
6
0
9
8
.
2
0
1
.
5
1
a
1
.
1
8
a
1
.
3
5
1
.
4
7
0
.
1
9
0
.
1
9
2
.
4
1
2
.
6
5
3
.
1
2
3
.
4
6
5
.
7
6
5
.
9
6
2
.
3
7
2
.
3
9
6
.
2
8
6
.
1
3
7
.
8
8
7
.
7
4
2
4
.
4
9
4
.
2
1
1
.
8
6
1
.
9
3
8
.
5
8
a
7
.
0
5
a
3
.
0
8
3
.
0
1
9
7
.
1
3
9
8
.
8
0
1
.
5
7
1
.
5
3
1
.
4
9
1
.
4
2
0
.
1
7
0
.
1
0
2
.
6
5
2
.
6
4
3
.
3
8
3
.
3
7
5
.
8
1
5
.
9
1
2
.
3
0
2
.
2
8
6
.
2
2
6
.
1
6
7
.
7
5
7
.
6
5
3
4
.
1
7
4
.
4
4
1
.
7
7
1
.
7
0
7
.
0
7
7
.
0
4
3
.
0
4
3
.
0
3
9
9
.
0
5
9
9
.
6
0
1
.
3
3
1
.
2
4
1
.
1
8
1
.
3
2
0
.
0
7
0
.
0
0
2
.
2
3
2
.
3
3
2
.
9
5
3
.
1
9
5
.
6
7
5
.
7
5
2
.
2
3
1
.
8
7
5
.
8
4
5
.
8
0
7
.
6
7
7
.
4
9
4
4
.
3
8
4
.
3
4
1
.
9
2
1
.
9
5
6
.
8
7
6
.
9
7
2
.
5
6
2
.
5
5
9
7
.
8
0
9
5
.
3
0
1
.
6
1
1
.
5
5
1
.
2
6
1
.
3
4
0
.
0
5
0
.
0
0
2
.
6
1
2
.
6
6
3
.
3
1
3
.
3
1
5
.
6
7
5
.
6
8
2
.
1
9
2
.
2
1
6
.
1
1
5
.
8
7
7
.
5
4
7
.
5
8
5
4
.
4
3
4
.
4
0
1
.
9
7
1
.
9
5
7
.
3
4
7
.
2
9
3
.
0
9
3
.
0
2
9
9
.
3
9
9
9
.
3
2
1
.
6
1
1
.
5
7
1
.
2
2
1
.
2
8
0
.
1
0
0
.
1
1
2
.
6
1
2
.
7
6
3
.
4
0
3
.
2
0
5
.
9
0
5
.
9
7
2
.
4
4
2
.
3
6
6
.
2
0
6
.
0
8
7
.
7
2
7
.
7
2
6
4
.
2
0
4
.
0
0
1
.
8
2
1
.
8
1
7
.
2
3
6
.
8
0
2
.
9
0
2
.
8
6
9
4
.
7
0
9
5
.
2
6
1
.
4
7
1
.
4
8
1
.
2
5
1
.
2
2
0
.
0
6
0
.
1
3
2
.
5
7
2
.
3
2
2
.
9
3
3
.
0
9
5
.
7
8
5
.
7
6
2
.
2
6
2
.
4
0
5
.
7
5
5
.
5
0
7
.
6
0
7
.
8
2
7
4
.
2
2
4
.
3
6
1
.
8
9
1
.
9
2
7
.
1
1
7
.
4
0
2
.
9
6
2
.
6
7
9
7
.
5
9
9
7
.
1
1
1
.
5
4
1
.
5
4
1
.
2
6
1
.
1
2
0
.
1
1
0
.
1
6
2
.
6
3
2
.
6
1
3
.
2
6
3
.
2
9
5
.
8
6
5
.
6
2
2
.
3
3
2
.
3
2
5
.
7
0
6
.
0
1
7
.
8
8
a
7
.
2
3
a
9
4
.
2
7
4
.
1
3
2
.
0
2
1
.
9
4
7
.
2
2
7
.
0
7
3
.
0
3
3
.
0
5
9
9
.
7
1
1
0
0
.
0
4
1
.
6
2
1
.
6
3
1
.
4
8
1
.
5
1
0
.
2
4
0
.
2
8
2
.
7
5
2
.
6
6
3
.
3
8
3
.
4
6
6
.
0
0
6
.
0
6
2
.
3
4
2
.
5
7
6
.
0
1
5
.
8
5
7
.
2
7
7
.
3
2
1
0
3
.
5
7
a
4
.
4
8
a
1
.
7
6
a
2
.
0
0
a
6
.
9
0
7
.
7
1
3
.
0
1
3
.
2
3
9
9
.
9
8
a
9
2
.
3
6
a
1
.
4
3
1
.
4
4
1
.
6
2
1
.
4
2
0
.
0
0
0
.
2
0
2
.
5
5
2
.
5
5
2
.
5
9
3
.
2
0
5
.
6
6
5
.
2
6
2
.
2
2
2
.
4
4
5
.
8
3
6
.
3
7
7
.
6
2
7
.
4
9
1
2
4
.
5
7
4
.
4
8
2
.
1
2
2
.
0
8
7
.
2
1
6
.
9
5
2
.
7
3
2
.
8
3
9
8
.
3
7
9
3
.
9
2
1
.
7
6
1
.
7
9
0
.
8
7
0
.
9
3
0
.
1
0
0
.
2
3
2
.
6
9
2
.
7
3
3
.
2
9
3
.
3
3
5
.
5
3
5
.
6
3
2
.
1
1
2
.
1
6
6
.
0
2
6
.
0
3
7
.
6
4
7
.
8
0
a
D
a
t
a
e
x
c
l
u
d
e
d
b
y
C
o
c
h
r
a
n

s
t
e
s
t
,
p
=
2
.
5
%
(
1
-
t
a
i
l
)
.
Table 5. Extractor type and laboratory ranking scores
for collaborating laboratories
Laboratory Extractor type Lab ranking score
1 Soxtec 2050 63.5
2 Soxtec HT 6 1043 55.5
3 Soxtec HT 6 1043 107.5
a
4 Soxtec 2050 91
5 Soxtec HT 6 1043 44.5
a
6
b
Soxtec 2050 107
7 Soxtec 2050 90
9
a
Soxtec HT 6 1043 48
10 Soxtec 2055 88
11
b
Soxtec 2055
b
12 Soxtec HT 6 1043 69
a
Results show bias using the lab ranking test.
b
XY plot revealed outlier results.
lets). This suggests that one of the significant sources of vari-
ability in the method is related to sampling the test portion.
The high fat supplement proved to be a challenge for the
collaborators. The fat (RSD
r
= 1.29%, RSD
R
=1.88%) melted
during the predry step, and foamed during extraction. Collab-
orators unfamiliar with this type of material may not have ob-
served these potential sources of error. As a result of the com-
ments received, better instructions have been incorporated
into the method describing how to handle these materials to
avoid incomplete recovery of crude fat.
Another challenge to the collaborators was the water-wash.
This was required for the texturized feed (RSD
r
=3.07%, RSD
R
=6.27%) and the feedlot concentrate pellets (RSD
r
= 5.80%,
RSD
R
=14.1%). Even though the current AOAC Official
Method 920.39 requires this step, many laboratories did not
have experience with it before the collaborative study, and this
is reflected in the high RSDs compared to materials that did not
require a water-wash. If materials which were water-washed
are removed from the statistics, the within-laboratory RSD
r
ranges from 1.23 to 4.96% for feed, forage, and cereal materi-
als, and among-laboratory RSD
R
ranged from 1.88 to 8.48%.
Therefore, based on the recovery studies, the water-wash step is
not necessary when hexanes are used as solvent, and only
serves to add error rather than improve recoveries.
HORRATvalues for dehydrated alfalfa, corn silage, mixed
bird seed, texturized feed, fat supplement, calf starter medi-
cated, calf feed medicated, meat meal/hulls mix, swine feed,
broiler starter, and high oil corn ranged from 0.71 to 1.93 and
are excellent. Two materials had HORRAT values >2.0: the
medicated goat feed (HORRAT = 2.26) and feedlot concen-
trate pellets (HORRAT = 3.67). The feedlot concentrate pel-
lets were challenging because of the water-wash step, and be-
cause the pellets had a low fat content, the RSD
R
represents
weighing differences among laboratories on the order of 8 mg
(on a weight of ~60 mg fat residue in an extraction cup weigh-
ing 25 000 mg for an aluminum extraction cup to 60 000 mg
for a glass extraction cup). Although the HORRAT of 3.67
sounds excessive, under closer scrutiny, it is easily accounted
for and may be improved upon by eliminating the water-wash
step. The HORRAT for medicated goat feed was slightly over
the desired 2.0, and not a real concern to the Study Directors.
One laboratory had an elevated Cochrans score, but escaped
removal by the Cochrans test. If this laboratory had been re-
moved, it would have lowered the HORRAT to 1.68.
906 THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003
Table 6. Study survey results
Lab Prewash Predry, 2 h Solvent
Instrument temp.
setting, C Soak, min Rinse, min
Postdry,
0.5 h Flow Comments
1 5,8,15,28 100C EM Science HX 0297-1 30183 135 20 40 100C Steady steam Yes
2 5,8,15,28 102C Fisher H303-4 013117 130 20 40 100C Rapid drip None
3 5,8,15,28 102C JT Baker 9308-33 unknown 140 20 40 100C None
4 5,8,15,28 102C Fisher H303-4 013115 135 20 40 100C 35 drips/s None
5 5,8,15,28 102C Fisher H291s-4 010762 145 20 40 100C 45 drips/s (rapid
drips)
Mix sand with
water-wash
6B 5,8,15,28 102C Riedel de Haen 15671 135 20 40 100C
Merck 104368 K29342968 124
7 5,8,15,28 102C Lab Scan C2536 1179/1 155 20 40 100C 2 + 3 drops/s Yes
9 5,8,15,28 102C Fisher (UN1208) H291-2c 001933 170 20 40 100C Distinct drop 35
drips/s
Yes
10 5,8,15,28 102C Fisher E291-4 011065 155 20 40 100C 4 drips/s None
11 None 103C Fisher H303-4 996668 102 25 30 100C Yes
12 5,8,15,28 102C Fisher E292-4 011382 185 20 40 100C Rapid drip Yes
Table 7. Recovery of crude fat, hexanes extract, from
ground corn spiked with urea and glucose
Urea spike, % Recovery, %
0.00 100.0
2.20 101.1
5.03 101.4
10.23 100.3
15.07 100.3
15.22 91.3
a
, 98.2
b
Glucose spike, % Recovery, %
0.00 100.0
2.16 96.5
4.99 98.1
10.30 98.1
15.07 98.4
15.04 90.4
a
, 98.2
b
a
With water-wash compared to unwashed control.
b
With water-wash compared to washed control.
Crude fat is an empirical method, i.e., it falls into the
Type 1 Codex Alimentarius Commissions scheme of defini-
tion of method types. A defining method is a method which
determines a value that can only be arrived at in terms of the
method per se (7). As discussed by Horwitz et al. (8),
gravimetric methods have limits of detection and precision
that are related to weighing error, and methods by which the
analyte is empirically defined are traditionally prone to greater
inherent variability than methods that are calibrated against a
reference standard. The HORRAT values observed in this
study are favorable with those reported by Horwitz et al. for
fat. In 105 fat assays with a concentration range of 370%,
they report an average RSD
R
of 14%, and a 90% confidence
interval for RSD
R
% and HORRAT of 0.565 and 0.212, re-
spectively. In this study with a concentration range of
1.599%, the average RSD
R
was 5.4%and RSD
R
ranged from
1.9 to 14% and HORRAT values from 0.71 to 3.7. The fact
that HORRAT values for 2 materials are >2.0 does not invali-
date the method. In fact, the method appears to be as good as
or better than those currently available as Official Methods,
and the hexanes appear to be as good as or better than diethyl
ether as a solvent.
Because the analyte is defined by the method, it is crucial to
emphasize that fat methods must be followed exactly. As
THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003 907
3.30
3.40
3.50
3.60
3.70
3.80
3.90
4.00
4.10
0 5 10 15 20
Urea spike as % of test portion
C
r
u
d
e
f
a
t
,
h
e
x
a
n
e
s
e
x
t
r
a
c
t
,
%
Crude fat, % Crude fat, % after water wash Trendline
Figure 1. Effect of urea on crude fat (hexanes extract)
recovery.
3.30
3.40
3.50
3.60
3.70
3.80
3.90
4.00
4.10
0 5 10 15 20
Glucose spike as % of test portion
C
r
u
d
e
f
a
t
,
h
e
x
a
n
e
s
e
x
t
r
a
c
t
,
%
Crude fat, % Crude fat, % after water wash Trendline
Figure 2. Effect of glucose on crude fat (hexanes
extract) recovery.
Table 8. Recovery of crude fat, hexanes extract, without predry step on collaborative study samples
Material
Moisture, % average after
rehydration
Crude fat, % study mean
(from Table 2003.06)
Crude fat, % hexanes
extraction, no predry Recovery, %
Dehydrated alfalfa 15.5 4.34 4.74 109
Corn silage 12.9 1.91 2.02 106
Mixed bird seed 15.4 7.15 6.87 96
Texturized feed 17.3 2.91 3.41 117
Fat supplement 6.70 97.77 96.76 99
Medicated goat feed 14.3 1.54 1.61 102
Feedlot concentrate pellets 27.8 1.30 1.00 77
Cellulose (blank) 18.5 0.12 0.16 134
Calf starter medicated 15.0 2.58 2.67 103
Calf feed medicated 13.4 3.23 3.33 103
Meat meal/hulls mix 24.5 5.76 6.28 109
Swine feed 17.8 2.29 2.08 91
Broiler starter 17.3 5.99 5.78 97
High oil corn 19.0 7.63 7.73 101
Average recovery 103
Horwitz et al. (8) concluded, analysts attempt to improve a
method of analysis by shortening times and eliminating what
appear to be purposeless steps. This was certainly observed in
this study when some collaborators had to be convinced of the
necessity to perform the predry step, which they felt was su-
perfluous or not cost effective. These steps have obviously
been eliminated in many laboratories, and are a source of vari-
ability normally associated with the method.
Recommendations
On the basis of this study, the Study Directors recommend
that the method for Crude Fat, Hexanes Extraction, in Feed,
Cereal Grain, and Forage (Randall/Soxtec/Submersion
Method) be adopted as Official First Action. Based on the
RSD of the cellulose blank, it is recommended that values be-
low 0.5% crude fat be reported as <0.5%.
Acknowledgments
We thank Brian Steinlicht (South Dakota State University,
Olson Biochemistry Laboratories, Brookings, SD) for assis-
tance with preparing, splitting, and sealing familiarization,
and study samples; Foss North America for providing extrac-
tion thimbles to collaborators and for supporting the study
with shipment of study materials; Fisher Scientific for provid-
ing some solvents for the solvent comparability testing. We
also thank the following collaborators for their participation in
this study:
Wayne Adcock, Alabama Department of Agriculture and
Industries State Chemical Laboratory, Auburn, AL
Roy Coffin, Soil & Feed Testing Laboratory, Department
of Agriculture and Forestry, Charlottetown, PEI, Canada
Amy T. DeBaker, Matthew Gramse, and Fatthieh Shafiee,
Wisconsin Department of Agriculture, Trade, and Consumer
Protection, Madison, WI
Pat Hogan, Sure-Tech Laboratory, Indianapolis, IN
John MacDonald and TomKnese, Ralston Analytical Lab-
oratories, St. Louis, MO
Jrgen Mller, Elisabet Frankenius, and Eva Bogren, Foss
Tecator AB, Hoganas, Sweden
Randy Royle, Servi-Tech Laboratories, Dodge City, KS
Marcy Russell and Jeff Boedigheimer, Foss North Amer-
ica, Eden Prairie, MN
Brian Shreve and Kelli Conway, Servi-Tech Laboratories,
Hastings, NE
Joel Sieh, CN-Labs, Courtland, MN
Misti Spann and Mick Watts, Kansas Department of Agri-
culture Laboratory, Topeka, KS
Erika Tpler, Schsischer Landeskontrollverband,
Niederwiesa, Germany
References
(1) Randall, E.L. (1974) J. Assoc. Off. Anal. Chem. 57,
11651168
(2) Official Methods of Analysis (2000) 17th Ed., AOAC IN-
TERNATIONAL, Gaithersburg, MD, Official Method
991.36
(3) Youden, W.J., & Steiner, E.H. (1975) Statistical Manual of
the AOAC, AOAC, Arlington, VA
(4) Official Methods of Analysis (2000) 17th Ed., AOAC IN-
TERNATIONAL, Gaithersburg, MD, Official Method
920.39
(5) Methods of Analysis for Nutrient Labeling (1993) D.M.
Sullivan & D.E. Carpenter (Eds), AOAC, Arlington, VA,
p. 94
(6) Pomeranz, Y., & Meloan, C.E. (1994) Food Analysis Theory
and Practice, 3rd Ed., Chapman & Hall, New York, NY,
pp 684692
(7) Codex Alimentarius Commission (1986) Procedural Manual,
6th Ed., Food and Agricultural Organization, Rome, Italy,
p. 139
(8) Horwitz, W., Albert, R., Deutsch, M.J., & Thompson, J.N.
(1990) J. Assoc. Off. Anal. Chem. 73, 661680
908 THIEX ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003