Chapter 4
Evaluation of the Rapid, High-Temperature Extraction
of Feeds, Foods, and Oilseeds by the ANKOM
XT20
Fat
Analyzer to Determine Crude Fat Content
R.J. Komarek, A.R. Komarek, and B. Layton
ANKOM Technology Corporation, Macedon, NY 14502
Abstract
The process of extraction for the quantitative separation of fat/oil is the basis for
the majority of official methods. The extraction process, which separates the sam-
ple into two fractions, permits two approaches to quantitative measurement. The
analysis can be performed by either weighing the fat/oil fraction directly, or indi-
rectly by measuring the loss of weight due to extraction. Acceleration of the
extraction process has been achieved by elevating the temperature of the solvent.
This chapter discusses a recently developed primary method called the Filter Bag
Technique (FBT). This technique utilizes temperatures of up to twice the boiling
point of petroleum ether to accelerate extraction. High sample throughputs are
accomplished by batch processing of samples encapsulated in filter media formed
in the shape of a bag. The extraction is performed automatically in an ANKOM
XT20
Fat Analyzer, an instrument that can process 20 samples in 20–60 min. The fat/oil
percentage is calculated indirectly from the loss of weight from the sample in the
filter bag. Various studies related to the extraction and gravimetric measurements
of these fractions are discussed in this chapter for both the conventional method
and the FBT. The accuracy of the FBT depends on effective predrying and proper
weighing of the sample. Studies of the conventional method suggest that samples
containing polyunsaturated fatty acids are sensitive to oxidation particularly during
the solvent evaporation step when the oil is heated in the presence of oxygen.
Various studies of the ruggedness of the FBT indicate that the method is not sensi-
tive to small changes in analytical conditions. The ruggedness of the method was

Both fat and oil represent the fraction of lipids generally associated with triacyl-
glycerides and compounds of similar solubility in nonpolar solvents. In this chap-
ter, the terms “fat” and “oil” will be used interchangeably.
The quantitative analysis of “Oil” as it is termed by American Oil Chemists’
Society (AOCS) (1) or “Crude Fat,” as designated by Association of Official
Analytical Chemists (AOAC) (2), is based on separating the fat/oil from the sam-
ple matrix by extraction with nonpolar solvents. The amount of oil is determined
either by directly weighing the extracted oil (Direct Method, AOAC Method
920.39a) or by measuring the loss of weight from the sample (Indirect Method,
AOAC Method 920.39b, 948.22a). This process is described in the flow diagram in
Figure 4.1. Each step in the process affects the accuracy and precision of the analy-
sis. There are several critical drying, weighing, extraction, and evaporation steps.
The process terminates with two fractions, i.e., the residue extracted by the solvent,
for which the percentage can be calculated directly, and that portion of the sample
not soluble in the solvent for which the percentage can be calculated indirectly.
Because both values can be determined on the same sample, their agreement veri-
fies the accuracy of the analysis.
Nonpolar solvents such as diethyl ether, petroleum ether, and hexane dissolve fats
and oils and leave behind proteins, carbohydrates, and other compounds insoluble in
these solvents. This fractionation is the basis for most of the “Official” analytical
methods established by AOCS, AOAC, International Organization for Standardization
(ISO) (3), German Fat Science Society (DGF) (4), and Federation of Oils, Seeds and
Fats Associations (FOSFA) (5). These methods utilize either the Soxhlet extraction
apparatus, developed by Franz Von Soxhlet (6) in 1939, the Butt-type apparatus (2),
or the Goldfisch apparatus (7). All of these methods boil the solvent and utilize the
condensed solvent to extract the sample. The Soxhlet apparatus allows the sample
chamber to fill and periodically siphon off into the boiling flask; the others simply
allow the condensed solvent to pass through the sample as the solvent is refluxed. The
Copyright © 2004 AOCS Press
sample is therefore extracted with solvent at a temperature below the boiling point of

background of the extraction process and the evaluation of the precision (repro-
ducibility among different laboratories in a collaborative study), accuracy (compar-
ison with standard methods), and ruggedness of the FBT in laboratory and interlab-
oratory collaborative studies.
Materials and Methods
Conventional Method. The Goldfisch Method, conducted on a Labconco
Goldfisch Fat Extraction Apparatus, was used in a number of studies as the con-
ventional standard for comparison with the FBT (7). The apparatus functions
essentially the same as the Butt-type apparatus, continually refluxing solvent over
the sample during the extraction. The method can follow both paths, i.e., direct
analysis and indirect analysis of fat/oil (Fig. 4.1). Extractions were performed over
a 4- to 5-h period and the solvent was partially evaporated and recovered in a glass
beaker. In earlier studies, the residual solvent (~10 mL) was evaporated above the
hot plate on a holder in the apparatus. In subsequent studies, with sensitive sam-
ples, the residual solvent was evaporated on a steam bath under nitrogen. The
analysis was conducted by weighing the sample in a tared thimble, drying the sam-
ple at 100°C for 3 h, and weighing it at ambient temperature from a desiccant
pouch. The thimbles in these studies were made from the hydrophobic filter medi-
um used for the filter bags. Typical cellulose thimbles are very hydroscopic and are
difficult to weigh. The thimbles containing the samples were inserted into the
apparatus and a tared glass beaker with 50 mL of petroleum ether was attached to
each reflux unit. The cycle was started by turning on the hot plate. When the
extraction was completed and the solvent evaporated, both the residual sample in
the thimble and the fat/oil in the beaker were dried at 100ºC for 30 min, cooled to
room temperature in a desiccator, and weighed. Both direct and indirect analyses
were performed on the same sample as a check for accuracy.
Filter Bag Technique. The FBT follows the path in Figure 4.1 of the indirect
analysis and was performed in the XT20 (9). The sample was weighed in the filter
bag, heat sealed, dried at 100ºC for ~3 h, cooled in a desiccant pouch, and
weighed. Samples (n = 20) were placed in a carousel in the extraction chamber.

Sample Preparation. The objective of sample preparation is to provide a sample
that accurately represents the “population” being studied and sufficiently disrupts
the matrix to permit more efficient extraction. Meat samples were ground to a uni-
form consistency with a food processor and mixed thoroughly. For shipping conve-
nience and sample uniformity, the meats in the international collaborative studies
were dried for 3 h at 100°C and then ground in a cyclone mill to pass through a 2-
mm screen. The feed samples were ground in a cyclone mill to pass through a 1-mm
screen and mixed thoroughly. The food samples were processed with a food proces-
sor or cyclone mill to produce a representative sample of uniform consistency.
Soybean samples were first dried at 130°C for 30 min and then ground in a cyclone
mill to pass through a 1-mm screen. Other oilseeds were ground in a cyclone mill to
pass through a 1- or 2-mm screen, depending on the level of screen occlusion.
The effects of grinding were demonstrated in a study with soybeans by pro-
cessing them three ways. In the first treatment, soybeans were ground through a 2-
mm screen and extracted. In the second treatment they were processed according to
the AOCS procedure (13) by first heating the soybeans in a 130°C oven for 30 min
and then grinding through a 1-mm screen followed by an extraction. The third
Copyright © 2004 AOCS Press
treatment involved regrinding the soybean samples from the second treatment
through a 1-mm screen and then extracting a second time.
Conventional Method Weighing Procedures. The weighing procedure is critical
to the gravimetric analysis of fats/oils. Accuracy of the analytical balance was veri-
fied and checked each day that weighing was performed. Accurate weighing of dried
samples requires rapid processing directly from a desiccating environment, limiting
exposure to moist ambient air. The glass beakers used in the conventional method
were hydroscopic and can, under certain circumstances, carry a significant static
charge. The effect of static charge was investigated in an experiment with samples of
a pig diet. Samples were extracted for 4 h with petroleum ether, and the residual oil in
beakers from six replicates was dried at 100°C for 30 min. After equilibration to
room temperature in a desiccator, the beakers were weighed. The oil was then trans-

The remaining treatments were dried in the oven at 100°C for 30 min. When samples
were removed from the oven they were equilibrated to room temperature in a desicca-
tor purged with nitrogen. The vacuum in the desiccator was returned to atmospheric
pressure with nitrogen.
Oil recovered from treatments 1, 5, and 6 was analyzed by thin-layer chromatog-
raphy (TLC). Samples were chromatographed on silica gel plates with methylene chlo-
ride and visualized with bromo thymol blue (14). This procedure separates the sterols,
triacylglycerides, and the less polar fractions.
FBT Predrying. Before extraction, all samples were dried at 100°C for 3 h for
both the conventional and the FBT methods. It is particularly important to remove
the residual moisture from samples analyzed by the FBT because the moisture is
removed during the extraction process, causing erroneously inflated values. A
study was made of the effects of predrying on ground beef, a high-energy horse
diet, corn, soybeans, and a pig diet for different periods of time and at different
temperatures. Samples were weighed in a filter bag and dried at 100, 105, and
110°C. The samples were analyzed at intervals of 30 min up to 180 min and each
treatment was replicated three times.
FBT Sample Size. The effect of sample size (1.00, 0.50, and 0.25 g) on the precision
of the analysis of six corn and three soybean samples was investigated in a study with
the FBT. The samples, analyzed in triplicate, were finely ground and had a uniform
consistency. Because of the sensitivity of the analytical balance (capable of weighing
to 0.1 mg) and the relatively small tare weight of the filter bags (0.5 g), it was expect-
ed that weighing errors would be minimized and the variance associated with this
study would be related to sample handling and sample homogeneity.
FBT Extraction Temperature. Because elevated solvent temperatures enhance the
extraction kinetics, the effects of extractions at three temperatures, 85, 90, and 95°C
were studied. Samples were extracted in 15-min intervals over a 60-min period. The
FBT analyses were conducted in triplicate on ground beef, soybeans, potato chips, and
a high-energy horse diet.
FBT Postextraction Drying. After extraction and solvent evaporation in the

were ground beef, cheese curls, soybeans, corn, and a horse diet. The conventional
analysis was performed by ANKOM Technology.
FBT Collaborative Study. A collaborative study, performed in conjunction with
AOCS, was designed to evaluate the precision and accuracy of the FBT with a wide
variety of samples that represented foods, feeds, meats, and oilseeds. Samples (n = 28)
were sent to 12 laboratories in the United States, Canada, and Europe in the form of 56
blind duplicates. Each laboratory was given a detailed protocol and had an opportunity
to become familiar with the method in a preliminary study. These samples were also
analyzed by three AOCS Certified Laboratories using the relevant official methods.
Results and Discussion
Reusing Solvent. The results of the solvent fractionation study of petroleum ether
(Fig. 4.2) indicated that the majority of the solvent (~70%) distilled in the range of
36–40°C with no other fraction >8%. The distributions of all of the fractions were sim-
ilar for both recycled and purchased solvent. This study indicates that petroleum ether
can be recycled without significantly changing the distribution of the solvent compo-
nents.
Sample Matrix Disruption. Fats and oils that are not hindered by the sample
matrix or by various types of binding rapidly dissolve in fat solvents. Oils trapped
Copyright © 2004 AOCS Press
in plant cell matrices are particularly difficult to extract due to the cell wall. This
microstructure can act as a semipermeable membrane where larger molecules have
limited access to exit the structure even though the smaller solvent molecule can
penetrate the structure. Plant matrices are difficult to disrupt on a cellular basis,
and this has led to the development of extensive grinding procedures. The grinding
and regrinding procedures required in the AOCS and FOSFA methods for certain
oilseed samples attest to the difficulty of preparing these samples for analysis. The
grinding study with soybeans illustrates the problem of sample preparation for
complete extraction of the oil (Fig. 4.3). The drying of the whole soybean at 130°C
for 60 min before grinding improved the yield by ~3%, whereas regrinding after
extraction improved the recovery by another 2%. In both treatments, it would be

Treatment 1 Treatment 2 Treatment 3
Copyright © 2004 AOCS Press
Oxidation. During a series of experiments with the conventional method, it was
found that for certain samples, such as hot dogs, ground beef, and potato chips, the
direct measurements of fat (the weights of the fat recovered) were in good agree-
ment with the indirect measurements (weight lost due to extraction) (Fig. 4.5). By
contrast, Figure 4.5 shows that the direct measurements of fat/oil were consider-
ably higher than the indirect measurements in oats, corn, soybeans, and a pig diet.
The distinguishing characteristics of this group include their plant origin and higher
concentrations of polyunsaturated fatty acids compared with the meat and potato chips
group. Similar studies with corn and oats also showed higher values for the direct
compared with the indirect analysis when solvent was evaporated on the Labconco
holder. Oxidation increases the weight of the oil (16), thereby increasing the direct
measurement of the oil. The extracted sample is not subject to the same effect, and no
distortion of the indirect measurement would be expected due to oxidation.
In the experiment designed to investigate variables in the method that would
enhance or avoid oxidation (Fig. 4.6), the indirect measurements of the oil content
were in excellent agreement across all six treatments. This was not true for the
direct measurement of the oil. Incremental changes in the time the oil was boiled in
Fig. 4.4. The effect of static charge on glass beakers was examined in five samples of
a pig diet by first weighing the fat in the beaker and then transferring the fat to alu-
minum pans and weighing the fat again.
Copyright © 2004 AOCS Press
the solvent during the reflux resulted in slight but inconclusive increases in the
direct value (Treatments 1–3). In Treatment 5, the solvent was evaporated using
the Labconco holder, which positions the beaker above the heater and allows the
temperature of the oil to rise above 100°C. The direct measurement of oil yielded a
value that was 4% higher that the indirect value. In Treatment 6, in which the sol-
vent was evaporated on the hot plate in the Labconco, the oil was subjected to tem-
peratures of 200°C for ~1 min. This resulted in a direct value that was lower than

Post Drying
Solvent Evap.
Reflux Time
Treatment
Copyright © 2004 AOCS Press
care has to be taken to avoid oxidation when measuring the oil fraction directly,
particularly with plant samples containing significant quantities of polyunsaturates.
Sample Predrying. During the refinement of the FBT, critical steps in the protocol
were investigated and optimized. The requirements of predrying were investigated for
a variety of sample types. A ground beef sample provides an example (Fig. 4.8) of the
relationship of moisture removal and the fat percentage. The percentage of dry matter
decreased for the first 120 min and then leveled off. The percentage of fat followed
the same pattern starting off high and leveling off after 120 min. The moisture that
was not removed in the oven was removed during the extraction and resulted in ele-
vated fat values. The three oven temperatures (100, 105, and 110°C) produced similar
results with ground beef. The same experiment with a high-energy horse diet (Fig.
4.9) indicated that the lipids in this diet were sensitive to temperature and that time in
the oven increased the effect. The horse diet, starting with <10% moisture, reached a
plateau in fat percentage in 60 min and maintained that plateau up to 150 min of
predrying at 100 and 105ºC. When the horse diet was heated at 110°C, a plateau in
the fat values was reached at 30 min and then declined exponentially after 60 min.
Fig. 4.7. TLC chromatogram showing the
separation of oil samples with different
heat treatments. Sample 1 was analyzed
under the mildest conditions; Sample 5
was heated on the Labconco holder and
Sample 6 was heated directly on the hot
plate. Triacylglycerides migrated to an R
f
of 0.45 and suspected oxidation degrada-

FBT accelerated the extraction of fats/oils and completed their removal in as little
as 15 min of extraction time for some samples (Table 4.1). This study examined
the effect of temperature and extraction time. Similar extraction rates were found
when the solvent was heated to 90 and 95°C. When the temperature was lowered
to 85°C there were some indications that the oil values were slightly depressed
(~0.5%) in soybean meal for shorter extraction periods. Generally, temperatures
of 90°C are effective for rapid removal of fats/oils for most samples.
Postextraction Drying. The studies that investigated the postdrying removal of
moisture and solvent residue showed that only a short drying time was required.
The study comparing 10 samples weighed directly after extraction and again after
10 min in the oven indicated a relative weight loss of 0.7% (SD 0.5). The weight
did not change with the second 10-min oven treatment, indicating that as little as
10 min in an oven at 100°C was sufficient for many samples. In the second study
in which samples were dried for up to 80 min, the longer drying period did not
change the fat value. A postdrying time of 30 min was chosen to ensure an effec-
tive drying period for all samples.
100°C
105°C
110°C
Time (min)
Fig. 4.9. The effect of time and temperature on the percentage of fat in a horse diet
during oven predrying (n = 3).
Copyright © 2004 AOCS Press
FBT Sample Size. A sufficiently large sample size is required to ensure proper
sample representation. In Table 4.2, the study of the effect of sample size of corn
and soybean on the precision of the FBT analysis showed that 0.5-g samples were
no more variable (SD 0.119) than 1-g samples (SD 0.111). Fat values for 0.25-g
samples were more variable (SD 0.278) with more than twice the SD of the 1 or
0.5-g sample. The fat value, however, was a reasonable estimate of the percentage
of fat for the 0.25-g sample. The variability imparted on the analysis by sample

SD 0.06 0.02 0.08 0.06 0.06 SD 0.37 0.33 0.27 0.42 0.35
avg. 24.3 24.4 24.2 24.1 avg. 32.8 32.6 32.7 32.8
SD 0.03 0.07 0.03 0.04 SD 0.37 0.25 0.22 0.46
a
n = 4.
Copyright © 2004 AOCS Press
the high level (Cap), except for soybeans (Table 4.3). Variations in sample size,
predrying time, predrying temperature, extraction time, extraction temperature,
postdrying time, and postdrying temperature had no effect on the FBT analysis. A
broad range of sample types was used to examine the effect of the method on different
sample matrices. The FBT analyses of meat (beef, chicken, and hot dogs), food (pota-
to chips), and feed samples (dog food, cattle feed, corn and poultry feed) were not
sensitive to the variables studied and proved to be very rugged for these samples.
Some of the variables had a significant effect on soybeans. There were indications of
sensitivity to sample size, extraction time, and temperature. This may well be due to
the difficulty in disrupting the ridged structures in the plant matrix. This again empha-
sizes that grinding of oilseeds is a critical step in the accurate determination of oil. The
automation of the extraction process by the XT20 contributed to the ruggedness of the
method by removing technician involvement.
Comparison of the FBT with the Conventional Method. The FBT in the intra-
laboratory study was found to be accurate and precise compared with the conven-
tional method of fat/oil analysis (Table 4.4). Regression analysis indicated an R
2
of
0.9996 between the FBT and the conventional method. The regression plot (Fig.
4.11) illustrates the excellent correlation between the two methods and indicates
that there is no bias between the methods. The regression line essentially passes
through the origin with a slope of one (Y = 1.001X – 0.046). Theoretically, a bias
should not be present when both methods use the same solvent and the extraction
conditions (time and temperature) are sufficient to complete the extraction. The

3 Predry temperature (°C) 98 102 22.26 22.31 0.05
4 Extraction time (min) 25 35 22.32 22.25 –0.07
5 Extraction temperature (°C) 89 94 22.32 22.26 –-0.06
6 Postdry time (min) 25 35 22.28 22.30 0.02
7 Predry time (°C) 98 102 22.32 22.26 –0.06
Significant at the 5% level diff = 0.31
Chicken thighs, cooked Hot dogs Corn
No. lc Cap diff lc Cap diff lc Cap diff
1 Sample size 34.01 34.03 0.02 45.61 45.71 0.10 14.31 14.20 –0.11
2 Predry time 33.99 34.04 0.05 45.69 45.63 –0.06 14.20 14.32 0.12
3 Predry temperature 34.05 33.99 –0.05 45.67 45.64 –0.03 14.30 14.21 –0.09
4 Extraction time 34.04 34.00 –0.04 45.60 45.71 0.11 14.31 14.20 –0.11
5 Extraction temperature 34.05 33.99 –0.06 45.64 45.67 0.03 14.25 14.26 0.01
6 Postdry time 4.02 34.02 0.00 45.66 45.65 –0.01 14.31 14.20 –0.11
7 Predry time 33.96 34.08 0.13 45.63 45.69 0.06 14.23 14.29 0.06
Significant at the 5% level diff = 0.16 0.42 0.27
Copyright © 2004 AOCS Press
Soybeans
b
Potato chips Cattle feed
No. lc Cap diff lc Cap diff lc Cap diff
1 Sample size 22.04 21.31 –0.73 36.44 36.45 0.01 3.31 3.24 –0.07
2 Predry time 21.58 21.77 0.18 36.46 36.43 –0.02 3.29 3.25 –0.04
3 Predry temperature 21.62 21.72 0.10 36.45 36.44 –0.01 3.31 3.23 –0.08
4 Extraction time 21.47 21.88 0.41 36.43 36.46 0.04 3.26 3.29 0.03
5 Extraction temperature 21.49 21.86 0.38 36.42 36.47 0.05 3.24 3.30 0.06
6 Postdry time 21.68 21.67 0.00 36.46 36.43 –0.04 3.23 3.32 0.09
7 Predry time 21.75 21.60 –0.15 36.43 36.46 0.04 3.25 3.30 0.05
Significant at the 5% level diff = 0.28 0.42 0.27
Poultry feed Dog food Ground beef

there was excellent agreement among laboratories analyzing the samples with the
FBT. The reproducibility between laboratories was S
R
= 0.43 and the repeatability
TABLE 4.4
Evaluation of the Precision and Relative Accuracy of the Filter Bag Technique
Compared with the Conventional Method
a
Conventional FBT
Sample % Fat/oil SD % Fat/oil SD
Rice hulls 0.3 0.07 0.2 0.08
Soybean meal 1.4 0.01 1.7 0.05
Pig starter 1.8 0.05 1.9 0.11
Chick grower 2.2 0.10 2.2 0.10
Cattle feed 2.7 0.10 2.8 0.08
Corn 3.0 0.07 3.5 0.12
Chicken breast 3.2 0.07 3.1 0.05
Blueberry muffin 4.6 0.41 4.7 0.39
Oatmeal 5.9 0.08 5.7 0.21
Brownie mix 8.8 0.07 8.5 0.15
Turkey 8.9 0.11 8.7 0.07
Fish meal 9.9 0.07 9.8 0.16
Ham 10.6 0.03 10.9 0.11
Soybeans 21.3 0.08 21.0 0.44
Horse feed 22.1 0.18 22.2 0.05
Tortilla chips 24.2 0.22 24.2 0.26
Ground beef 28.4 0.16 28.6 0.23
Chicken thighs 29.1 0.09 29.2 0.13
Sausage 36.4 0.35 36.7 0.62
Safflower 40.4 0.22 39.5 0.20

Fig. 4.11. Regression analysis of the accuracy of the FBT analysis of fat/oil relative to
the conventional method in 22 samples each replicated five times.
Copyright © 2004 AOCS Press
TABLE 4.5
Comparison of the Conventional Method with the Filter Bag Technique Performed in 13 Collaborating Laboratories on 5 Samples
Ground beef Cheese curls Soybeans Whole corn Horse feed
Filter bag method Average SD Average SD Average SD Average SD Average SD
Laboratory #
1 22.7 0.19 41.4 0.25 20.7 0.12 3.2 0.11 21.7 0.20
2 22.4 0.20 41.4 0.18 21.0 0.19 3.3 0.16 21.7 0.16
3 21.4 0.60 40.3 0.55 20.8 0.56 3.2 0.03 21.8 0.08
4 23.0 0.16 41.7 0.17 21.3 0.33 3.7 0.27 21.9 0.13
5 22.1 0.25 41.0 0.44 19.6 0.43 3.0 0.69 21.1 0.36
6 21.8 0.25 39.4 0.54 20.8 0.15 3.1 0.11 21.3 0.10
7 22.8 0.24 41.6 0.46 21.1 0.13 3.3 0.10 22.1 0.08
8 22.4 0.45 41.6 0.25 21.0 0.46 3.1 0.38 21.7 0.12
9 22.7 0.25 41.5 0.09 22.0 0.07 3.7 0.15 22.3 0.16
10 22.8 0.04 42.0 0.17 20.5 0.24 3.1 0.19 21.4 0.32
11 22.5 0.23 41.1 0.20 20.0 0.36 3.3 0.15 22.1 0.25
12 21.9 0.53 40.4 0.27 20.9 0.35 2.3 0.18 21.8 0.17
13 22.7 0.16 41.9 0.18 21.8 0.17 4.2 0.22 22.3 0.16
S
r
0.31 0.32 0.31 0.27 0.20
S
R
0.56 0.80 0.72 0.50 0.41
Average 22.4 41.2 20.9 3.3 21.8
Conventional method average 23.0 0.28 40.5 1.00 20.7 0.08 3.4 0.07 22.2 0.47
a

2 5.4 9.0 20.7 37.9 1.6 3.1 3.3 3.4 5.7 2.3 6.4 22.6 2.1 20.1
11 1 6.4 8.9 21.4 41.1 1.8 3.3 3.2 3.0 5.5 3.0 6.6 22.6 2.5 19.8
2 6.4 9.3 22.2 40.5 1.7 3.1 3.1 3.1 5.5 2.8 6.3 22.9 2.5 20.7
12 1 6.2 8.3 20.5 38.3 1.9 3.2 8.3 4.6 5.2 2.1 6.3 22.7 2.7 20.3
2 6.1 8.2 19.6 38.3 2.1 4.3 3.7 3.0 5.5 4.8 6.9 22.6 1.8 20.0
(Continued)
Copyright © 2004 AOCS Press