Sei sulla pagina 1di 24

MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO.

6, 2002 1217

AGRICULTURAL MATERIALS

Gravimetric Determination of Amylase-Treated Neutral


Detergent Fiber in Feeds with Refluxing in Beakers or Crucibles:
Collaborative Study
DAVID R. MERTENS
U.S. Department of Agriculture, Agricultural Research Service, U.S. Dairy Forage Research Center, 1925 Linden Dr West,
Madison, WI 53706-1108

Collaborators: M. Allen; J. Carmany; J. Clegg; A. Davidowicz; M. Drouches; K. Frank; D. Gambin; M. Garkie;


B. Gildemeister; D. Jeffress; C-S. Jeon; D. Jones; D. Kaplan; G-N. Kim; S. Kobata; D. Main; X. Moua; B. Paul;
J. Robertson; D. Taysom; N. Thiex; J. Williams; M. Wolf

As an important constituent of animal feeds, fiber and among-laboratory reproducibility standard de-
represents the portion of feeds that is bulky and viation (sR) varied from 0.37 to 2.24. The HORRAT
difficult to digest. The neutral detergent fiber (NDF) was <2 for all materials except feed materials con-
method, developed over 30 years ago, is the taining >10% fat. However, standard deviations of
method of choice for measuring total fiber in for- repeatability and reproducibility for feeds with
ages and other feeds. Several modifications that >10% fat were similar to those of other materials. It
were made to improve its general applicability to is recommended that the aNDF method be ac-
all feeds and others developed in individual labora- cepted for Official First Action status.
tories often resulted in variability among laborato-
ries in measuring NDF. The amylase-treated NDF
(aNDF) method, therefore, was developed as an ac- iber is nutritionally important because it represents the
curate and precise method of measuring total in-
soluble fiber in feeds. A collaborative study was
conducted to evaluate the repeatability and
F organic portion of feeds and foods that is the most diffi-
cult to digest; nonfiber fractions of feeds are easily and
almost completely digestible by most animals. There is a need
reproducibility of the aNDF method over the full for a rapid and simple assay to determine the total insoluble fi-
range of animal feed materials. Twelve laboratories ber content of feeds. Originally fiber was related to the cell
representing research, feed company, regulatory, wall fraction of plants; however, because fiber occurs in all
and commercial feed testing laboratories analyzed feeds, fiber methods must be applicable to all feeds and foods.
11 materials as blind duplicates. The materials rep- Although neutral detergent (ND) extraction solubilizes some
resented feed matrixes, including animal products; material that is not digested by mammalian enzymes, neutral
high-protein, high-fat, and high-pectin feeds; oil detergent soluble fiber is often fermented by bacteria in the gut
seeds; grains; heated by-product feeds; and le- of animals and is digested. Thus, neutral detergent fiber
gume and grass hays and silages. Materials se- (NDF) is the insoluble fiber in feeds that is either indigestible
lected varied in chemical composition and con- or slowly digested, and occupies space in the digestive tract of
tained 090% aNDF, 116% ash, 120% crude fat, animals.
140% crude protein, and 050% starch. Cor- The NDF method was originally developed by Van Soest
recting results for changes in blanks and reporting
and Wine (1) for the analysis of total fiber in forages, and nu-
results as ash-free aNDF organic matter (aNDFom)
improved the repeatability and reproducibility of merous modifications have been proposed to extend its appli-
results when aNDF was <25%. The within-labora- cation to grains, concentrated feeds, and human foods (25).
tory repeatability standard deviation (sr) for per- Most of these modifications were based on conceptual im-
centage aNDFom in feeds varied from 0.21 to 1.82 provement in total fiber analysis without ruggedness testing to
determine their suitability for all feed matrixes or practical ap-
Submitted for publication July 2002. plication as a routine method. Unfortunately, this resulted in
The recommendation was approved by the Methods Committee on fiber methods that differed, yet were all called NDF. The diffi-
Feeds and Fertilizers and Agricultural Related Products as First Action. See
Official Methods Program Actions, (2002) Inside Laboratory culty in extracting and washing fibrous residues in some mate-
Management, September/October issue. rials and the variety in modifications of the method have led to
Corresponding authors e-mail: davem@dfrc.wisc.edu.
the perception that NDF is difficult to measure precisely.
1218 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002

Collaborative Study fore, developed a visual assessment method for determining


the amount of any heat-stable amylase source that is active un-
Optimization and Ruggedness of NDF Methods
der the conditions used in aNDF analysis, and this is included
The critical steps in the NDF and amylase-treated NDF as part of the method.
(aNDF) procedures were investigated by the original develop- (7) Washing method.Rapid rinsing of fibers removes
ers, who proposed modifications, and the Study Director, who only surface contamination of detergent and soluble matter
developed the method evaluated in this study. The following and results in high fiber values. Research by the Study Direc-
factors were evaluated: tors laboratory indicated that multiple soakings of fibrous
(1) Time of refluxing.Sixty min of refluxing at boiling residues are required to obtain consistent results and uncon-
temperatures achieves asymptotic extraction for most materi- taminated fibrous residues.
als and was selected as an optimum compromise between
(8) Reporting fiber as organic matter.Ashing fiber resi-
analysis time and complete extraction (1). It should not vary
dues and reporting results as aNDF organic matter (aNDFom)
more than 5 min.
eliminates some differences in results associated with inade-
(2) Neutral detergent concentration.Sodium lauryl sul-
fate is used at near saturation concentrations to maximize quate washing of fibrous residues. In addition, it allows nutri-
solubilization of nonfiber with a minimal volume of liquid tionists to estimate nonfibrous carbohydrates more accurately
that must be filtered (1). by difference (NFC = 100 crude protein crude fat ash
(3) Neutral detergent pH.Phosphate and borate buffers aNDFom), because a portion of ash is not double-subtracted
are used to maintain pH near 7.0 because alkaline or acid con- as it would be if aNDF were used in the calculation. Calcula-
ditions can solubilize fiber. The Study Director observed that tions in the proposed method allow results to be reported to
laboratories were not confirming that pH was near 7.0 and meet the needs of the user.
conducted a test to evaluate the effects of pH variation. Re- (9) Sample amount and ratio to ND solution.The origi-
sults indicated that the pH should be maintained within the nal NDF method used 100 mL ND with 1 g sample (9).
range of 6.957.05 (6). Halving the amount of sample reduced aNDF by about 1 per-
(4) Triethylene glycol.The original NDF method used centage unit. To reduce filtering problems, the same ratio of
ethylene glycol monoethylether to remove nonfiber soluble sample to ND was maintained, but sample and ND solution
material from concentrated feeds. This compound is a poten- amounts were reduced to 0.5 g in 50 mL.
tial mutagen and was replaced with triethylene glycol. Re-
(10) Sample grinding.In most near infrared reflectance
search confirmed that changing to triethylene glycol did not
spectroscopy, samples are ground through a cyclone mill with
alter results (J.B. Robertson, Cornell University, Ithaca, NY,
1 mm screen. Discrepancies in fiber analysis were observed
personal communication, 1988).
when these samples were analyzed by chemical methods. Re-
(5) Sodium sulfite.Sodium sulfite was included to re-
move proteinaceous material from fiber residues. When search by the Study Director confirmed that particle size of the
heat-stable amylase was included as a part of the NDF method, sample and aNDF results were affected by both screen size and
sodium sulfite was removed because it has the potential to de- grinder type. Although samples that are ground finer are ex-
stroy phenolic compounds (2). However, research by the Study tracted more completely by ND, they are often more difficult to
Director indicated that sodium sulfite is critical for removal of filter. Thus, the proposed method specifies that materials must be
proteinaceous matter in heated or cooked feeds, and it was rein- ground through a 1 mm screen with a cutter mill or equivalent.
troduced in the method that is being evaluated (7). (11) Sample drying.Artifact fiber can be created during
(6) Heat-stable a-amylase.Starch was not completely sample preparation by drying materials at temperatures
removed by the original NDF method. Numerous modifica- >60C (10). Therefore, the method to be evaluated specifies
tions of the method used various amylases to remove starch that wet samples be dried at temperatures <60C.
before or during ND extraction. Both the amylase source and (12) Modifications for problem samples.Samples that
the method of using the enzyme alter NDF values, and pre-ex- are high in starch, pectin, fat, or animal proteins are difficult to
traction by incubation has the potential to solubilize fiber. filter. Modifications to the basic method have overcome the
Heating samples and gelatinizing the starch may improve am-
difficulties inherent in these types of samples (11).
ylase effectiveness in removing starch contamination. Many
amylase solutions are crude extracts that contain enzymes that As indicated by Horwitz et al. (12), empirical methods of
can degrade fiber and must be deactivated by near-boiling analysis, such as fiber methods, must be followed exactly. This
temperatures. Alternative methods of using heat-stable suggests that the various modifications of the original NDF
amylases have been evaluated by the Study Director and oth- method should have different names because they are different
ers (4, 8). The approach used in the proposed method uses 2 measures of fiber. The proposed fiber method is called aNDF to
additions of heat-stable amylase: one in ND after initiation of distinguish it from the original method and its modifications.
boiling, and one in the first residue washing step. The objective of this research was to evaluate the repeatability
Unfortunately enzyme sources changed or were discontin- and reproducibility of the aNDF method in a collaborative
ued and new specifications for the type and amount of amy- study involving laboratories with different missions and using
lase to use had to be rediscovered. The Study Director, there- materials representing the full scope of feed ingredients.
MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1219

Table 1. Characteristics of collaborating laboratories filtration vessels, balances, and drying ovens. Collaborators
Amylase were provided with test samples, heat-stable -amylase, and
Lab No. Type Method apparatus source chemicals, but were required to mix and standardize the neces-
sary reagents. They were asked to respond to several question-
0 Research (Study Director) Reflux beakercrucible A
naires about equipment and routine procedures in their labora-
1 Research Reflux beakercrucible B
tories, provide results of standardizing the amylase working
solution, analyze 5 familiarization samples, and provide aNDF
2 Research (dropped out)
results for 11 collaborative study materials and one blank ana-
3 Research Reflux beakercrucible A
lyzed as blind duplicates. Collaborators were asked to perform
4 Regulatory Reflux in crucible A
singles analyses of each sample, to report all weights to 4 signif-
5 Regulatory Reflux beakercrucible B
icant digits, and to submit data on an as-is basis.
6 Testing Reflux in crucible B
7 Feed company Reflux beakercrucible B Materials
8 Testing Reflux beakercrucible A
9 Testing Reflux beakercrucible A There are no officially recognized reference materials for
10 Testing (dropped out) fiber, and it is not possible to prepare spiked samples. To de-
11 Feed company Reflux beakercrucible B termine the general applicability of the aNDF method, study
12 Feed company Reflux beakercrucible B materials were selected to represent the full range of fiber con-
13 Feed company Reflux in crucible A centrations and types of animal feeds (Table 2). In addition,
14 Regulatory Reflux beakercrucible B materials were included that required the use of each of the
modifications described in the method. Materials representing
feed matrixes from animal products; high-protein, high-fat,
and high-pectin feeds; oil seeds; grains; heated by-product
feeds; legume forages; and grass forages (hays and silages)
Collaborating Laboratories were included in the study. A mixed dairy feed that contained
a mixture of animal and plant supplements with added fat was
The 14 laboratories which agreed to participate in the study used to represent a variety of feed ingredients.
represented 3 feed company laboratories, 5 commercial feed Materials selected for the collaborative study encompassed
testing laboratories, 3 feed regulatory laboratories, and 4 re- a wide range of aNDF concentrations in feeds, varying from 0
search laboratories. Two feed testing and one research labora- to 90%. The materials also varied in concentration of ash
tory failed to complete the study; the Study Directors labora- (116%), fat as measured by ether extract (120%), crude pro-
tory was substituted for the missing research laboratory and a tein (140%), and starch (050%). The diverse chemical com-
volunteer feed company laboratory was substituted for one of position of the materials used as blind duplicates permitted
the feed testing laboratories (Table 1). Participants received evaluation of most, if not all, possible interactions of feed
no compensation. Each laboratory provided reflux apparatus, composition on aNDF analysis. Blanks were included to eval-

Table 2. Materials used in the study, representing full range of fiber concentrations and types of animal feedsa
Material Description DM, % Ash, % EE, % CP, % Starch, % NFC, % ADF, % aNDFom, %

Alfalfa silage Forage, fermented legume 91.0 9.7 3.0 20.1 0.2 20.1 37.4 38.1
Brewers grains Concentrate, heated by-product 92.6 5.1 6.3 24.7 2.8 9.2 20.0 47.2
Citrus and beet pulp Concentrate, pectic by-product 91.4 7.8 3.9 7.3 1.3 45.7 25.4 26.7
Corn grain with cob Concentrate, containing starch 92.4 1.3 4.5 6.4 50.4 59.6 10.6 20.6
Corn silage Forage, grass with starch 92.3 3.8 5.0 6.5 33.4 41.5 22.3 35.4
Corn stalks Forage, high-fiber grass 92.0 7.4 1.0 4.1 0.2 11.2 46.5 68.2
Dairy mixed feed Concentrate, mixture with fat 92.6 16.1 13.5 39.8 3.9 11.3 7.4 11.9
Grass hay Forage, temperate grass 92.0 8.6 5.4 8.5 0.5 14.1 38.1 55.5
Milk replacer Concentrate, animal fat 93.5 7.0 17.6 18.6 0.6 50.2 0.3 0.1
Roasted soybeans Concentrate, plant fat 93.9 5.2 18.0 38.2 2.6 19.5 6.2 13.0
Sawdust Reference, high cellulose 92.5 1.6 0.8 0.6 0.6 0.4 77.6 89.1
Blank No material

a
Dry matter (DM), ash, ether extract (EE), crude protein (CP), starch, nonfibrous carbohydrates (NFC = DM ash EE CP aNDFom), acid
detergent fiber (ADF), and amylase-treated neutral detergent fiber organic matter (aNDFom) of materials used in the collaborative study.
1220 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002

Table 3. Homogeneity of aNDFom (blank-corrected) for 4 sample sets for each material analyzeda

Material n Mean, % sr sR RSDr, % RSDR, % 2.8 sr 2.8 sR HORRAT

Alfalfa silage 4 38.1 0.50 0.52 1.31 1.37 1.40 1.46 0.59
Brewers grains 4 47.2 0.46 0.59 0.98 1.24 1.30 1.64 0.55
Citrus and beet pulp 4 26.7 0.36 0.50 1.33 1.87 1.00 1.40 0.77
Corn grain with cob 4 20.6 0.65 0.65 3.17 3.17 1.83 1.83 1.25
Corn silage 4 35.4 0.18 0.34 0.51 0.97 0.51 0.96 0.42
Corn stalks 4 68.2 0.82 0.84 1.20 1.23 2.29 2.35 0.58
Dairy mixed feed 4 11.9 0.30 0.31 2.52 2.57 0.84 0.86 0.93
Grass hay 4 55.5 0.22 0.42 0.39 0.76 0.61 1.18 0.35
Milk replacer 4 0.1 0.29 0.32 373.29 409.78 0.81 0.89 69.70
Roasted soybeans 4 13.0 0.36 0.53 2.76 4.07 1.01 1.49 1.50
Sawdust 4 89.1 0.56 0.60 0.63 0.68 1.58 1.69 0.33
Blank, g 4 0.0009 0.0012 0.0015 0.0035 0.0042

a
sr = Standard deviation of replicated analyses within sample set; sR = standard deviation within and among sample sets; RSDr = repeatability
relative standard deviation; RSDR = reproducibility relative standard deviation.

uate weighing technique and determine if blank-correction Familiarization samples were sent to each collaborator in a
improved precision. preliminary study to acquaint them with the method and eval-
Single batches of homogenous materials were obtained from uate their ability to handle difficult samples. Familiarization
commercial feed suppliers or the U.S. Dairy Forage Research materials included alfalfa silage (AS), barley hay (BH), CG,
Center (Madison, WI). Materials were dried as needed at 55C meat and bone meal (MB), and RS. The latter 3 materials were
and were ground through a 1 mm screen with a Wiley cutter selected to represent materials that were extremely difficult to
mill, and thoroughly mixed. Forty sets of samples containing analyze for aNDF.
about 2 g of each of the 11 materials were prepared. Samples Heat-stable amylase is a critical reagent in the aNDF
were measured into glass vials that were sealed in mois- method. To verify that alternative sources of amylase can be
ture-proof pouches for storage until shipment. All samples were standardized and used, 3 amylase stock solutions from
numbered in the order in which they were to be analyzed. 2 sources were standardized by each collaborator during the
familiarization study, and half of the laboratories were as-
Homogeneity of sample sets was verified by selecting 4 sets
signed to use one of the 2 sources during the collaborative
of samples at random for each material and analyzing them in
study (Table 4).
duplicate in the Study Directors laboratory to provide an esti-
mate of random variation within and among sets of samples. Statistical Analyses
The results of the homogeneity study were analyzed by using The experiment was designed as a randomized complete
the AOAC spreadsheet statistical software recommended for block design using blind duplicates of 11 unidentified materi-
collaborative studies to determine among-sample set instead of als. Although feeds were not identified, milk replacer, dairy
among-laboratory variation (Table 3). For most materials, the feed, and roasted soybeans were labeled to indicate >10% fat
variation between duplicate analyses within a sample set (sr) in the first blind duplicate, but not the second. Only one ana-
was similar to the variation between samples sets (sR). This lyst from each laboratory was asked to submit results, and
analysis allows variability in homogeneity of sample sets to be blind duplicates were analyzed on the same day within the
compared directly with the variation among collaborating labo-
ratories. Except for milk replacer (MR), which had an aNDF
concentration near zero, the Horwitz ratio (HORRAT; 12) of Table 4. Sources of stock heat-stable a-amylases
observed versus expected relative standard deviations (RSDs) evaluated and used in the study
among results was below the acceptable threshold of 2.0. For
Amylase type Source Dilution
most samples, the HORRAT was between 0.33 and 0.58.
Within the Study Directors laboratory, the aNDF variation was
similar to that of other AOAC Official Methods. The HORRAT Taka-Therm L-340 ANKOM Technology Full strength
Corp. (Fairport, NY)
values for roasted soybeans (RS), corn grain (CG), dairy mixed
feed (DM), and citrus and beet pulp (CB) were between 0.80 Termamyl 120 L Novo Nordisk Biochem Full strength;
(Raleigh, NC) one-half strength
and 1.50, suggesting that these materials were less homoge-
neous or more difficult to analyze for aNDF.
MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1221

same run. Block was defined as first replicate, second repli- Y = Material + Lab + Block + error (3)
cate, and rerun analyses using modifications. In the first repli-
cate, all laboratories were asked to analyze one duplicate of
Predicted results were compared with observed results, and
each of the samples in a fixed order of decreasing filtration
the replicate with the largest residual deviation was identified
ease. In the second replicate, one of 12 random orders of anal-
as the outlier. Laboratory ranking scores (after removal of in-
ysis for the samples was assigned to each laboratory. Ordering
dividual replicate outliers) were used to identify laboratories
samples in the first replicate was used to determine the effect
that were outliers across all materials according to
of a difficult-to-filter sample on the subsequently analyzed
Wernimont (13). All outlying results were checked for calcu-
samples. The statistical model for analysis of variance
lation and data entry errors, and answers to the questionnaire
(ANOVA) was:
were scrutinized to determine if the collaborators procedure
differed from the aNDF method that was being evaluated.
Y = + Material + Lab + Block + Material*Lab + error (1)
Within laboratory repeatability (sr), among laboratory sys-
tematic variation (sL), and total reproducibility variation (sR;
where = the mean, Material = variation due to materials, Lab =
where sR2 = sr2 + sL2) associated with a single analysis were
variation among laboratories, Block = variation due to replicate
determined by approaches described by Youden and
or rerun, Material*Lab = variation due to Material by Laboratory
Steiner (14) and Wernimont (13) with spreadsheet statistical
interaction, and error = random variation within laboratory.
software recommended by AOAC. Statistical results were de-
Block was included in the model to test the effect of order of
termined for individual materials. In addition, ANOVA using
analysis and was removed from the model when not significant.
the interaction model (Model 1) was used to determine if re-
There are several procedures in the aNDF method that may
sults from all materials or classes of materials (forage, concen-
not be followed exactly among laboratories. For example, identi-
trates with <10% fat and concentrates with >10% fat) could be
fying the proper level of amylase to use when determining aNDF
pooled. If the interaction term in Model 1 was not significant,
is an integral part of the method because the source and level of
it was assumed that the method obtained consistent results
enzyme is not specified in the method. A portion of the collabo-
among materials, and pooled precision statistics were calcu-
rative study was designed to evaluate the effects of amylase
lated according to Wernimont (13).
source (A or B) and dose (independently determined by each col-
laborator). The effect of source and dose of amylase was evalu- True values cannot be determined for fiber because the em-
ated on all materials and only on materials containing significant pirical method defines the fraction. Therefore, there is no direct
starch [corn silage (CS) and CG] using the ANOVA model: measure of systematic bias (accuracy) for fiber, and consensus
values derived from the collaborative study were used as the
point of reference. As suggested by Youden (15), variability in
Y = + Material + Amylase + Dose + Amylase*Dose + error (2)
systematic bias was equated with among-laboratory variability
(sL) that does not include within-laboratory variability. Repeat-
where Amylase = variation due to source of amylase, Dose
ability (sr) is the within-laboratory precision that was calculated
and Amylase*Dose = variation explained by linear regression
from variation between blind duplicates within laboratories.
of amylase dose for each source, and all other variables as de-
scribed previously. Reproducibility (sR) is the total variation associated with a
Because aNDF is an empirical method it is important to single analysis, which is the sum of systematic bias and re-
evaluate the effects of factors that are left to the discretion of peatability, i.e., sR2 = sr2 + sL2. Repeatability within a labora-
laboratories and may differ among them. Each collaborator tory is expected to be 1/22/3 of the total reproducibility varia-
submitted answers to a questionnaire about factors that might tion among laboratories. Although specific criteria for the
affect results. In addition to amylase source and dose, collabo- acceptability of a fiber method cannot be imposed, it is ex-
rators were asked to describe for each result the apparatus, fil- pected that the variation among laboratories will be a function
tering aid, and pre-extraction technique used, and to describe of aNDF concentration. The ratio (HORRAT) of the observed
the temperature of washing water and the methods for residue reproducibility relative standard deviation (RSDR) divided by
washing and weighing. This information was evaluated using the expected RSD from the Horwitz equation (16) was used to
appropriate ANOVA models. evaluate the acceptability of the method. A HORRAT <2 was
used as an indicator of acceptability (12).
Initially, outlying results were detected with spreadsheet
statistical software provided by AOAC for a blind duplicates
design using a 2.5% significance level. The Cochran test iden- AOAC Official Method 2002.04
tifies replicate results within a laboratory that are suspect, and Amylase-Treated Neutral Detergent Fiber in Feeds
the Grubb test determines if the average result of laboratories Using Refluxing in Beakers or Crucibles
First Action 2002
deviates from those of all laboratories. Cycles of Cochran, sin-
gle Grubb, and pair Grubb tests were used to identify outliers [This method is applicable for the determination of amy-
until no additional removal was necessary or no more than 2/9 lase-treated neutral detergent fiber (aNDF) in forages, grains,
of the laboratories were flagged. The ANOVA identified grain by-product feeds, animal by-product feeds, oil seeds,
which of the 2 blind replicates was the outlier by using a main oilseed meals and mixed feeds that range in concentration
effects ANOVA with no interaction terms: from 1.5 to 100%.]
1222 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002

Caution: Enzyme (amylase) preparations can cause allergic Although boiling ND solution dissolves most proteins, lipids,
reactions in hypersensitive individuals; avoid in- and nonfibrous carbohydrates, these constituents are nonviscous
haling aerosols or dusts. Cover eyes, nose, mouth, only in water that is near boiling temperature. Therefore, boiling
and exposed skin as needed to prevent irritation. water is needed for washing fibrous residues and removing
Sodium lauryl sulfate is mildly irritating to mucous nonfibrous matter. Because fiber is particulate matter, mass-ac-
membranes and may cause allergic reactions in hy- tion equilibration during soaking is needed to migrate ND and
persensitive individuals. Do not inhale or allow contaminating soluble matter from the interior of fiber particles.
material to contact skin or eyes. Wear a dust mask Fibrous residues must be soaked, instead of being simply rinsed,
and gloves while handling. Triethylene glycol is in near boiling water to remove nonfibrous matter.
harmful if inhaled, ingested, or absorbed through Boiling ND solution solubilizes lipids, and acetone soak-
the skin. Vapors or mists are irritating to the eyes, ing of the residue completes the extraction of lipids and pig-
mucous membranes, and upper respiratory tract. ments in most materials. However, excessive amounts of
Contact is irritating to the skin. Wear appropriate lipids in materials can complex with the detergent and reduce
clothing and mix in a hood or well-ventilated area. extraction efficiency. Because extraction of lipids by ND and
Acetone is a flammable solvent; do not use near acetone may be incomplete when feeds contain >10% lipid,
open flames. Use in an effective fume hood and do these materials are pre-extracted to ensure complete removal
not let vapors accumulate in work area. Avoid in- of lipid contamination from fiber. Pre-extract materials with
haling or contact with skin. Evaporate all traces of 510% lipid to minimize filtration difficulties and avoid
acetone before placing fiber residues into an oven. variable aNDF results.
Avoid skin contact with hot neutral detergent solu- B. Apparatus
tion and reagents. Maintain all electrical equipment
in suitable working order and ensure that it is prop- (a) Refluxing apparatus.Any conventional apparatus
erly grounded. suitable for crude fiber or acid-detergent fiber determinations.
Test samples should be extracted in 500 or 600 mL beakers
See Table 2002.04 for the results of the interlaboratory without spouts (Pyrex No. 1040, or equivalent) that are covered
study supporting acceptance of the method. with a round cold-water condenser to minimize evaporation.
Calibrate heating unit setting so that 50 mL water boils within
A. Principle 45 min when cold water condensers are used. Fibertec appara-
Fiber in feeds is a nutritionally defined entity that repre- tus 2010 or M6 (Foss North America, Eden Prairie, MN 55344)
sents the indigestible and slowly digesting fraction of feeds. can be used and should boil 50 mL water within 10 min.
Neutral detergent (ND) solution and heat-stable -amylase (b) Fritted-disk Gooch crucibles.Coarse porosity (pore
are used to dissolve easily digested proteins, lipids, sugars, size 4060 m) crucibles, high-form, 4050 mL capacity or
starches, and pectins in feeds, leaving a fibrous residue that is Fibertec USP2 (pore size 4090 m, 2628 mL capacity).
primarily cell wall components in plant materials (cellulose, Clean new crucibles and ash at 500C for 1 h. Clean crucibles
hemicellulose, and lignin) and indigestible nitrogenous matter after each use by ashing at 500C for 3 h, removing ash, in-
in animal products. aNDF is a gravimetric method that best es- verting in detergent solution, and sonicating for 710 min.
timates the total insoluble fiber, and is inversely related to di- Rinse crucibles in hot water, and soak in room temperature
gestibility and intake potential of a feed. ND soluble matter is water for at least 30 min. Back-flush crucibles by connecting
almost completely digested by most animals; however, the di- top of each crucible to a No. 9 stopper or a No. 9 stopper fit-
gestibility of aNDF is variable among feeds and is related to ted with a No. 4 filter adapter (Cat. No. 24035-087; VWR,
lignin and other constituents in fiber. West Chester, PA 19380, www.vwr.com) that has a port that
Sodium lauryl sulfate, an anionic detergent, and sodium sul- is connected to a trap and vacuum line. Back-flush each cruci-
fite are used to solubilize nitrogenous matter; EDTA is used to ble with water by rapidly plunging and removing the bottom
chelate calcium and enhance removal of pectins at boiling tem- of the crucible into water to create a vigorous rinsing action.
peratures; triethylene glycol helps remove some nonfibrous Occasionally test filtration rate as follows: Fill each cruci-
matter from concentrated feeds; disodium phosphate and so- ble with 50 mL distilled water (25 mL for Fibertec USP2 cru-
dium borate are used as buffers to maintain a neutral pH. cibles) and record time required to drain completely without
Hot ND solution has limited ability to solubilize starch; vacuum (should be 180 60 s for Gooch or 75 30 s for
therefore, a heat-stable amylase is used to hydrolyze starch to USP2). If <100 s (or <35 s for USP2), discard crucible. If
saccharides that can be easily removed from fiber by filtration. <120 s (or <45 s for USP2), check for cracks in fritted disk. If
Heat-stable amylases are used in hot solutions to inactivate filtration >240 s (or >105 s for USP2), clean crucibles with
potential contaminating enzymes in crude amylase extracts slow filtration rates using acid or alkaline cleaning solutions.
that might degrade fibrous constituents. To ensure that the am- (c) Vacuum filter manifold.Any apparatus similar to that
ylase activity is sufficient to remove most starch and to reduce described by Mertens et al. (17) or suitable alternative that al-
filtering difficulties, the amount of any specific amylase lows adequate soaking of fibrous residues (Figure 2002.04A).
source needed to measure aNDF is determined under the con- Manifold should provide vacuum-tight seal with crucible to re-
ditions of the aNDF method. duce foam formation in vacuum lines. Use thick-walled vac-
MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1223

Table 2002.04. Interlaboratory study results for amylase-treated neutral detergent fiber in feeds
Feed Fiber Mean, % sr sR RSDr, % RSDR, % ra Rb

Forages aNDFc 52.2 0.84 1.10 1.61 2.10 2.36 3.08


d
Forages aNDFbc 52.2 0.84 1.08 1.61 2.08 2.36 3.03
Forages aNDFome 50.4 0.93 1.18 1.85 2.34 2.61 3.31
f
Forages aNDFombc 50.1 0.93 1.16 1.85 2.32 2.60 3.26
Concentrates <10% fat aNDF 33.5 1.18 1.42 3.51 4.25 3.29 3.98
Concentrates <10% fat aNDFbc 33.2 1.25 1.57 3.76 4.72 3.50 4.38
Concentrates <10% fat aNDFom 32.5 1.16 1.46 3.56 4.48 3.24 4.08
Concentrates <10% fat aNDFombc 32.2 1.14 1.47 3.55 4.56 3.20 4.11
Concentrates >10% fat aNDF 8.7 1.01 1.61 11.66 18.57 2.83 4.50
Concentrates >10% fat aNDFbc 8.7 0.79 1.24 9.06 14.26 2.21 3.47
Concentrates >10% fat aNDFom 8.8 0.58 1.20 6.54 13.68 1.61 3.37
Concentrates >10% fat aNDFombc 8.5 0.53 1.10 6.25 12.94 1.49 3.09
All materials aNDF 38.7 1.05 1.33 2.72 3.44 2.94 3.72
All materials aNDFbc 38.6 1.02 1.28 2.64 3.32 2.86 3.59
All materials aNDFom 37.7 1.02 1.28 2.70 3.39 2.85 3.58
All materials aNDFombc 37.4 1.00 1.24 2.67 3.32 2.80 3.48

a
2.8 sr.
b
2.8 sR.
c
Amylase-treated neutral detergent fiber (as-received basis).
d
Amylase-treated neutral detergent fiber, blank-corrected (as-received basis).
e
Amylase-treated neutral detergent fiber organic matter (as-received basis).
f
Amylase-treated neutral detergent fiber organic matter, blank-corrected (as-received basis).

uum tubing to connect manifold to trap (418 L) and vacuum tended nozzle with larger opening to adequately rinse particles
source. Place a vacuum reservoir (18 L) between trap and vac- into ND solution during extraction.)
uum source to ensure adequate vacuum capacity to remove
foam. Fibertec apparatus may be used for filtration. C. Reagents
(d) Boiling water supply.Use continuous boiling water
generator as described by Goering and Van Soest (9) or suit- (a) Sodium sulfite.Na2SO3 anhydrous.
able alternative (Figure 2002.04B). Apparatus must be capa- (b) Dried hominy corn (corn grits, raw).Obtained from
ble of supplying boiling water (>95C) in quantity sufficient food or grocery store. Must be raw or uncooked grits; pre-
for all residues to be washed at one time through a nozzle pro- cooked or instant grits are not acceptable. Grind grits to pass
ducing a fine stream (flow rate, 3540 mL/10 s; a 2.5 mL dis- 1 mm screen in a Wiley cutter mill.
posable plastic pipet tip makes an acceptable nozzle). A fine (c) Burkes iodine solution.2 g KI and 1 g I in 100 mL
nozzle minimizes water needed to transfer particles to cruci- water. Store in amber or opaque bottle.
ble, but provides water pressure needed to remove residues at- (d) Stock heat-stable a-amylase solution.Termamyl
tached to side of beaker. It is critical that water is boiling when 120 by Novozyme Corp. (Cat. No. A3403, Sigma-Aldrich, St.
added to crucibles, especially for products containing starches, Louis, MO 63178), Spezyme FRED by Genecor International
pectic substances, mucilages, or glycoproteins. (Fibertec users: (Cat. No. FAA, Ankom Technology Corp., Fairport, NY
Use syringe with cone-spray nozzle to rinse condensers and 14450) or equivalent liquid heat-stable -amylases or water
60 mL disposable syringe with 12 gauge needle 10 cm long to extracts of heat-stable -amylase powders.
dislodge any residues adhering to condensers.) (e) Working heat-stable a-amylase solution.Standard-
(e) Wash bottle for ND solution.Rinse down particles ize heat-stable -amylase stock solution or enzyme powder
that are carried up the side of beakers by foaming during extract so that 2 additions of 2 mL will remove raw corn starch
refluxing. Use fine nozzle to minimize addition of ND solu- from 0.5 g of corn hominy grits. Assay as follows:
tion during rinsing (flow rate, 1015 mL in 3 s with moderate (1) Weigh 0.5 g (0.005 g) ground, dried hominy corn into
pressure; unclipped nozzle on 500 mL Nalgene Unitary wash each of 6 beakers similar to those used to extract fiber residues
bottle or equivalent is acceptable). (Fibertec users: Use ex- described in B(a).
1224 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002

Figure 2002.04A. Schematic drawing of vacuum filter manifold: (A) side view; (B) top view; and (C) end view (ref. 17).
MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1225

(8) After solutions are 20 0.5C (may take 5 min or


more), remove beakers from tempering bath and arrange in or-
der of increasing enzyme doses on white background.
(9) Quickly add 0.5 mL Burkes iodine solution to beak-
ers, and mix.
(10) Do not look at beakers. After 90 s, view through solu-
tions from above and make a quick decision (within 30 s)
about color of each solution, using the following scale: Purple
= not adequate enzyme; pink-amber or amber = not adequate
enzyme; pale yellow = adequate enzyme.
Compare to blank. Brown tint of enzyme solution should
not be confused with pink-amber or amber. If color differ-
ences are unclear, place beakers in tempering bath for 5 min
and repeat steps (8)(10).
(11) After lowest dose that is pale yellow (V2) and next
lowest dose (V1) that is pink-amber or amber are identified us-
ing a geometric progression of doses, do final standardization
using the dose below the pink-amber one (V1) and a linear pro-
Figure 2002.04B. Water heating unit. Heating unit on
gression of doses (0.25*V1) between V1 and V2 (e.g., if
right consists of a 3 L round flask modified to contain 2
0.05 mL treatment is amber [V1] and 0.10 mL treatment is pale
standard taper female joints for cold water condensers
at the top and a glass column (6 cm diameter, 17 cm yellow [V2], use doses of 0.025, 0.05, 0.0625, 0.075, 0.0875,
length) at the bottom to accommodate a 1000 watt quartz and 0.10 mL) stock solution in the final standardization.
immersion heater, 455 mm long and 19 mm diameter (12) The lowest dose that is pale yellow (and exceeds next
(Cat. No. 16790-1L, Vycor sheathed heating element, highest inadequate dose with pink-amber or amber solution)
Corning Glass, Corning, NY, vycor@corning.com, or represents the volume of amylase stock solution or extract
equivalent), inserted through the original joint at top. (Vs) used to make amylase working solution.
Cold water enters through a port at the bottom of the (13) Record date, batch or lot of amylase, doses tested,
column and leaves through a port in the upper 2/3 of the amount of iodine solution used, and color and final tempera-
round flask. The tank at the left contains a float valve
ture (before iodine addition) of each dose in reagent log book.
adjusted to maintain a water level in the heating unit
about 3/4 full.
(14) Determine number (n) of test samples to be analyzed
in the following 5 days or less. To add 2 mL amylase working
solution twice for each test requires a total volume of amylase
working solution of (n) 4 mL. Mix (n) 2 (Vs) mL amy-
(2) Preheat calibrated reflux units, prepare ice bath for lase stock solution with (n) [4 2 (Vs)] mL water. Store
cooling beakers (must contain enough ice to maintain temper- working amylase solution in refrigerator using stoppered con-
ature below 1C), and prepare tempering bath [shallow pan tainer no longer than 5 days.
containing enough water at exactly 20C to exceed depth of (15) Confirm adequacy of amylase working solution by re-
solutions in beakers and maintain a final temperature between peating standardization procedure using 0.5 g corn grits with 0,
19.5 and 20.5C at the end of step (8)]. 2, and 4 mL working solution (each added at boiling and after
(3) Add 50 mL ND solution, (f), (do not add sodium sul- removal from refluxing apparatus). If there is no appreciable
fite), swirl beaker, and place on preheated refluxing apparatus difference in color between 2 and 4 mL working solution, then
at 1 min intervals. 2 additions of 2 mL are adequate for the aNDF method.
(4) After ND begins to boil (ca 5 min), add one of 6 doses Notes: Time and temperature are critical for proper assess-
stock or powder extract solution (geometric progression, e.g., ment of adequate amylase dose. The solutions must be 20C
0, 0.025, 0.05, 0.10, 0.20, and 0.40 mL; exact doses will de- and the decision about color must be made within 90120 s af-
pend on source of amylase) to beakers in ascending order. ter addition of iodine solution. Pink-amber or amber colors
fade quickly, and waiting longer than 120 s before making a
(5) Reflux for 10 min, remove at 1 min intervals; add sec-
decision will result in a dosage that is too low.
ond dose of amylase (matching the first), swirl, and rinse sides
Initial doses of stock solution or extract may have to be ad-
of beaker using minimum of room temperature ND solution.
justed, and the standardization rerun. If maximum dose
(6) React 60 s and filter through glass wool or 2 layers of (0.4 mL) results in purple or pink-amber color, extend the geo-
cheesecloth into 100 mL glass beaker. Prepare blank by add- metric progression of doses starting at 0.4 mL, and rerun ini-
ing 2 intermediate doses, e.g., 0.10 and 0.20 mL stock solu- tial standardization. If minimum dose (0.025 mL) is yellow,
tion, to 40 mL room temperature ND in 100 mL glass beaker. decrease the geometric progression to between 0 and
(7) Place beakers, except blank, in ice bath. Remove from 0.025 mL, and rerun initial standardization.
ice bath after 5 min (temperature of solutions should be ca Each new source or lot of enzyme should be standardized; a
21C) and place all beakers in tempering bath (20C). single lot that is being used over a period of time should be
1226 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002

checked every 6 months for activity. Excess enzyme is not ben- mean particle size of 220260 mm. A 1 mm screen in a cutter
eficial and can be detrimental. Concentrated enzyme solutions mill (Wiley) or 2 mm screen in cyclone or centrifugal mill is
are not recommended as working solutions because a 1 drop er- generally acceptable. Cyclone or centrifugal mills throw parti-
ror when dispensing enzyme contains significant activity that cles through the screen at an angle, which effectively reduces
can affect results. Many amylase extracts are crude mixtures the size of the aperture that particles pass through. On material
that may contain fibrolytic and proteolytic activities. Heat-sta- ground to finer particle size, fiber must be determined by
ble amylase solution must be used in hot liquids (>80C) to in- microfiber filter mats to minimize particle loss; however, re-
activate contaminating enzymes and minimize fiber loss. sults may have negative bias. Grinding segregates the
(f) Neutral-detergent (ND) solution.Measure 990 mL material, with highest fiber material passing out of grinder
water. Pour ca half of water into a flask, add 4.0 g NaOH, last. Do not discard material in grinder; combine it with mate-
14.6 g EDTA, 4.56 g dibasic sodium phosphate (Na2HPO4), rial in grinder receptacle. Mix ground material by placing on
6.81 g sodium borate decahydrate (Na2B4O710H2O), and mix creased sheet of paper (ca 40 40 cm). Lift corners of paper to
until dissolved (heat if necessary). Under a hood, add 30 g so- roll material to diagonal corner, turn paper 90, and roll; repeat
dium lauryl sulfate and of remaining measured water. Mix 10 times. Transfer material to suitable container.
until detergent is dissolved and add 10 mL triethylene glycol Ground test samples can be pre-extracted with acetone or
to suppress foam. Add remaining water and thoroughly mix. ether to remove fat, but do not dry at >60C. Products contain-
Verify that pH is between 6.95 and 7.05, and adjust with con- ing >5% fat should be pre-extracted; those with >10% fat must
centrated HCl or NaOH, as required. If pH is off by >0.5, dis- be pre-extracted to remove fat. To pre-extract with acetone,
card. Store ND solution at room temperature or, if cool storage weigh test portion into crucible, place on filtering manifold,
causes precipitation, warm to 25C, and mix before use. Re- extract with 4050 mL acetone 4 times (allow material to soak
cord date ND solution was prepared, pH measurements, and at least 5 min, and stir 3 times during each soaking), apply vac-
adjustments in reagent log book. uum to remove traces of acetone, air-dry for 1015 min, en-
Note: NaOH and EDTA can be replaced with 18.6 g disod- sure that all traces of acetone are removed, and transfer to re-
ium EDTA. flux beaker. Use same crucible to collect fiber residue for test
(g) Filter aid.(1) Silica sand.Sand, cristobalite, acid portion after ND extraction. If filtering aid is used, dry and
purified, 40200 mesh (Fluka, Buchs, Switzerland, Cat. No. weigh it with crucible, and then transfer it to a small beaker
84880 or equivalent). (2) Glass microfiber mats.4.25 cm before test sample is weighed into crucible. After pre-ex-
Whatman GF/D or equivalent. tracted residues have been transferred from the crucible into
(h) Crucible cleaning solutions.(1) Acid.Chromic the refluxing beaker, replace filtering aid in crucible before fil-
acid [see Definition of Terms (13)] or prepare from 2.5 L tering ND-extracted residues. [Fibertec users: Add silica sand
H2SO4 and one package Nochromix (Godax Laboratories, to USP2 crucible, dry and weigh it before adding test sample;
Inc., Takoma Park, MD 20912, www.godax.com), or equiva- then pre-extract with 2030 mL acetone 4 times (allow mate-
lent inorganic oxidizer. (2) Alkaline.Dissolve 5 g rial to soak at least 5 min and stir with back pressure 3 times)
Na2EDTA2H2O 50 g Na3PO4, and 200 g KOH in 1 L water. using cold filtration device. After removing all traces of ace-
Cleaning procedure.After testing filtration rate of cruci- tone, transfer crucible containing silica sand and pre-extracted
bles, place crucibles with slow filtration rates in a shallow test portion to the hot extraction device for ND extraction.]
glass or enamel pan and add ca 40 mL acid cleaning solution
to each crucible. Let acid cleaning solution filter through cru- E. Determination
cible and soak for 1 h. Rinse with water by back flushing and Dry at 105C for >4 h and weigh empty crucibles (hot di-
retest filtration rate. Clean crucibles with slow rates with the rectly from oven or room temperature after desiccation). Re-
alkaline cleaning solution. Place crucibles in a shallow pan cord empty crucible weight for test portions (We) or blanks
and add 50 mL alkaline cleaning solution and heat to (Be) to nearest 0.0001 g. Mix material thoroughly and weigh
7080C. Let alkaline cleaning solution filter through the cru- 0.5 (0.0500) g air-dry feed, or equivalent amount of wet ma-
cible and then back flush each crucible until it is one-half full terial into refluxing beaker. (Fibertec users: Weigh test
of solution. Let solution filter through crucible and back flush portion into dried and preweighed USP2 crucible.) Record fi-
crucible until it is one-half full. Let solution filter through cru- nal weight of test portion to nearest 0.0001 g (S). If results are
cible. Remove from alkaline solution and back flush each cru- to be reported on dry matter basis, weigh a second test portion
cible with water. After it has cooled, alkaline cleaning solution at the same time for dry matter determination. Include
can be saved and reused. Use alkaline cleaning solution spar- in-house reference sample and 2 blanks for first 2030 test
ingly because it dissolves glass and weakens the fritted disks samples in a run and add one reference and one blank for each
in crucibles. Retest crucible filtration rate; do not use those additional 2030 test samples.
with slow rates for aNDF analysis. Preheat calibrated reflux units. Add 0.5 (0.1) g sodium
sulfite using calibrated scoop (EKCO Housewares Pinch
D. Test Sample Preparation
measuring spoon, World Kitchen, Inc., Elmira, NY,
Dry wet laboratory samples at <60C to prevent creation of www.worldkitchen.com or equivalent) and 50 (5) mL ND to
artifact fiber. Fiber extraction is affected by particle size of test each refluxing beaker and swirl (critical for starchy feeds that
sample. Grind representative samples to obtain geometric stick to bottom during refluxing). (Fibertec users: Use back
MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1227

pressure to mix reagents.) Do not add ND and sodium sulfite but remove water to leave a damp or moist residue, before
to test portions more than 60 min before refluxing. Heat to adding acetone. Excessive evacuation clumps the residues and
boiling within 45 min, add 2 mL working amylase solution, makes acetone extraction difficult.
resuspend particles stuck to bottom or sides, and swirl. Vacuum residue dry, remove crucible from manifold, and
(Fibertec users: Use back pressure to mix amylase with ND air dry for 1060 min to remove acetone. Dry crucibles at
solution and test portion.) 105C for minimum of 8 h and weigh (Wf and Bf). Ignite cruci-
Reflux for 60 min at boiling temperature that creates vigorous ble and fiber in 500C furnace for 5 h or until C-free. Temper in
particle movement, but not excessive foaming that carries parti- 105C oven for at least 1 h and weigh (Wa and Ba). Weigh cru-
cles up the side of the beaker. Mixtures may foam vigorously for cibles containing fiber or ash residues (hot or desiccated to
12 min (do not reduce temperature of heating unit). Rinse sides room temperature) in same order as empty crucibles.
of beaker with minimum amount of ND using bottle with fine Modifications for specific types of test samples.(a) If
nozzle 510 min after amylase is added, and rinse as needed to extracted ND solution appears milky and opaque and filtration
resuspend particles on side of beaker (twice maximum). is slow during transfer of residues or after first water soaking,
Remove extracted mixture from heating unit and let parti- high starch is suspected. Add additional treatment with 2 mL
cles settle for 3060 s. Before transfer (Fibertec users: Before amylase during second water soaking. Shorten soaking times to
initial filtration), observe mixture to determine if lipid globules minimum to keep soaking solutions as hot as possible (>85C).
are present on surface or if solution is milky, which indicates (b) If residue clogs crucible during transfer and additional
that test sample should be rerun after acetone pre-extraction. amylase does not improve filtration, feed material may contain
Place Teflon stirring rod in crucible and preheat by adding proteinaceous, gum, or mucilage residues (meat products and
40 mL boiling water for 3060 s. (Fibertec users: Ignore.) some oil seed meals). Preheating crucible with boiling water is
Remove water with vacuum, and immediately decant top crucial for filtering these materials. The best filter aid for these
3040 mL of solution, keeping beaker inverted over crucible. materials is 1215 g (68 g for Fibertec USP2) of silica sand,
(Fibertec users: Ignore.) Use minimum vacuum to evacuate ex- C(g)(1). The gummy substances in these feed materials will
cess liquid, but close vacuum before residue becomes dry. stick to sand particles which prevents them from clogging the
Note: Excessive vacuum and evacuating to dryness cause some fritted disk and allows residues to be washed. All filter aids
residues to clog crucible and not wash properly. Rinse all unat- must be added to crucibles (including blanks) before initial
tached particles into crucible using fine stream of boiling water. weights are recorded.
(Fibertec users: Ignore.) Fill crucible half full with hot water. (c) If fiber residue has glossy, translucent sheen and filtra-
Add 2 mL working amylase solution and stir. (Fibertec users: tion becomes more difficult with each water soaking, pectic
Use back pressure to mix amylase in initial water soak.) substances are suspected. Preheat crucible with boiling water
React with amylase for minimum of 4560 s while scrap- and transfer residues as quickly as possible without settling
ing remaining particles from bottom and sides of reflux beaker when removed from reflux unit. Reduce all soaking times to
with rubber policeman. Evacuate amylase solution and trans- minimum to maintain temperature >85C to prevent cooling
fer any remaining residue from reflux beaker into crucible and jelling in crucible. Filtering aids may improve filtration (in
with 2030 mL boiling water. Two rinses are usually suffi- order of preference): 1215 g (68 g for Fibertec USP2) silica
cient. After transferring residues from beaker, fill crucible 3/4 sand, 0.25 g (0.15 g for Fibertec USP2) glass wool, and glass
full with boiling water and soak for 13 min. (Fibertec users: microfiber mats, C(g)(2).
Remove amylase-water soak after minimum reaction of 60 s. (d) If fat globules are observed floating on surface of ND or
Crucibles can be removed from hot to cold filtration unit for wash water and residue is difficult to filter, or if material is known
remaining hot water soaks for residues that are easy to filter. to contain >10% fat, pre-extract it with acetone or ether (see D).
This allows next set of test samples to begin ND extraction on (e) If material contains fine particles, flocculant precipitates,
hot filtration unit. Residues that are difficult to filter can be dirt (fine clay), or fecal matter, but not pectic substances or
washed on Fibertec heating unit with heat reduced to mini- starch, increase settling time to maximum of 2 min after removal
mize particle agitation.) from refluxing unit and use filter aid in crucible. Filter aids (in or-
Evacuate water, add 4050 mL boiling water, soak der of preference) include glass microfiber mats, ceramic fiber,
35 min, and repeat. If residues are difficult to filter after first 1215 g silica sand, and 0.25 g glass wool. Microfiber mats can
soak, add additional 2 mL working amylase solution. If resi- be gently scraped to renew surface during filtration.
dues appear translucent and become more difficult to filter (f) If all other modifications fail, reduce test portion amount
with each additional soaking, eliminate third water soak. If to 0.3 g and repeat analysis with filter aid in crucible. Reducing
plugged, crucible can be back-flushed by removing it from fil- test portion will magnify effects of weighing errors and increase
ter manifold and reinserting it. (Fibertec users: Use minimal variation in results. Sometimes reducing test portion amount
back pressure to open crucibles and improve filtration.) and increasing ND to 70100 mL is beneficial. If fiber is
Evacuate water, refill crucible with 4050 mL acetone, stir <1.5%, do not reduce test portion amount, and if filtration is not
to disperse particles, soak 35 min, and repeat, rinsing stir rod possible, report results as difficult to analyze, fiber <1.5%.
to remove attached fiber particles. (Fibertec users: Move cru- (g) Do not add acetone before all rinse water has been re-
cibles to cold extraction unit for acetone soaks.) Note: After moved. Although this occasionally improves filtering, it does
the last water soak, do not evacuate fiber residues to dryness, not remove detergent or detergent solubles from residues.
1228 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002

Adding acetone before water washing is complete will give lot in hot ND solution and adjust amylase working solution ac-
inflated fiber values. cordingly. Check activity of stock amylase every 6 months
during storage, and adjust working solutions accordingly.
F. Calculations
Include at least one in-house reference or quality control
Without blank correction: (QC) sample and 2 blanks for the first 2030 test samples in a
run, and add one QC and one blank for each additional 2030
% aNDF (as-is basis) = 100 (Wf We)/S test samples analyzed. Suitable QC materials include brewers
grains, grass hay, or corn silage that has been dried at <60C.
where aNDF is amylase-treated neutral detergent fiber and S
Each of these materials is sensitive to changes in reagents and
is g as-is test portion weight.
technique, but corn silage is preferred because it contains
% aNDF (DM basis) = 100 (Wf We)/(S*DM) starch and grass cell walls. Acceptable standard deviation
(SD) among repeated analyses of the reference material
where DM is (g oven-dried matter weight/g air-dried or wet should be 1.00% aNDF. Plot QC sample results on X-control
test portion weight). chart and examine chart for trends. Results outside of upper or
lower warning limits (2.00 standard deviations) are evidence
% aNDFom (as-is basis) = 100 (Wf Wa)/S of possible problems with the analytical system. Results out-
side of upper or lower control limits of 3.00 SDs indicate loss
where aNDFom is aNDF organic matter. of QC, and results of the run should be verified. Two consecu-
tive analyses falling on one side of the mean between warning
% aNDFom (DM basis) = 100 (Wf Wa)/(S*DM) limits and control limits also indicate loss of QC.
With blank correction (average all blanks before calculat- Include at least one set of duplicates in each run if single
ing results): determinations are being made. Duplicates should not be run
consecutively, but one replicate should be spaced at the
% aNDF (as-is basis) = 100 (Wf We Bf + Be)/S begining and end of the run. Acceptable differences among
duplicate analyses ranges from 1.50%-units for materials with
where aNDF is amylase-treated neutral detergent fiber, S is g <20% NDF to 3.00%-units for materials with >70% NDF.
as-is test portion weight, and Bf and Be are averages of all Change in weights of blank crucibles should be <0.0100 g
blanks in the run. after either ND extraction or ashing. If weights blank crucibles
change by >10 mg, or crucible weights after ashing are less
% aNDF (DM basis) = 100 (Wf We Bf + Be)/(S*DM)
than empty crucible weights, suspect inadequate cleaning of
where DM is (g oven-dried matter weight/g air-dried or wet crucibles or problems with weighing technique.
test portion weight). Ref.: J. AOAC Int. 85, 12211228(2002)

% aNDFom (as-is basis) = 100 (Wf Wa Bf + Ba)/S Results and Discussion

where aNDFom is aNDF organic matter. Familiarization Study

% aNDFom (DM basis) = 100 (Wf Wa Bf + Ba)/(S*DM) In the familiarization study, collaborators were provided
6 materials with a mean value and acceptable range for each.
Results should be reported to nearest 0.1%, and results The mean value and acceptable range were determined in the
<1.5% aNDF or aNDFom should be reported as aNDF Study Directors laboratory based on duplicate or triplicate
<1.5% or aNDFom <1.5%. analyses of 4 randomly selected sets of materials (Table 5).
When fiber is <25%, blank correction is required. Except for MB and RS, variation among sample sets was less
Crucible weights after ashing (Wa) that are less than empty than replication variation within sets, as indicated by the ob-
crucible weights (We) indicate that crucibles lost weight dur- servation that sR was equal to sr. For RS there was a small, but
ing the aNDF procedure. The crucible weight after ashing insignificant variation among sample sets. The variation
(Wa) is the most accurate estimate of the correct crucible among samples sets of MB was about 40% of the
weight for calculating fiber and must be used to calculate reproducibility standard deviation (RSDR), indicating a lack
aNDFom. In this circumstance, aNDFom is a more accurate of homogeneity among sample sets. The large sr and
estimate of fiber than is aNDF. HORRAT for MB and RS indicate that samples of these mate-
If results for in-house reference or quality control samples rials were less homogeneous and more difficult to analyze for
are outside of control limits, do not report results; re-analyze aNDF than other familiarization materials. Collaborators were
test samples. asked to analyze the materials using the aNDF method and to
repeat analyses using modifications given in the method until
G. Quality Assurance
they achieved acceptable results. The collaborators results in
Maintain log of reagent preparation and amylase standard- Table 5 are the best 2 or 3 results selected from all analyses
ization. Check pH of each batch or lot of ND solution and ad- they performed. However, several participants did not follow
just as needed. Determine activity of each amylase source and instructions and only provided one set of duplicate results.
MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1229

Table 5. Statistical data for 4 random sample sets of materials analyzed in the Study Directors laboratory and the
best 2 or 3 results for collaborator laboratories on materials in the familiarization study
Sample IDa n Mean, % sr sR RSDr, % RSDR, % rb Rc HORRATd

SDBH 4 54.3 0.78 0.78 1.44 1.44 2.20 2.20 0.66


SDAS 4 43.4 0.36 0.36 0.82 0.82 0.99 0.99 0.36
SDCS 4 38.9 0.20 0.20 0.52 0.52 0.57 0.57 0.23
SDCG 4 8.0 0.18 0.18 2.29 2.29 0.51 0.51 0.78
SDMB 4 19.5 1.50 2.42 7.70 12.39 4.20 6.76 4.84
SDRS 4 19.5 0.81 0.88 4.13 4.49 2.26 2.46 1.76
ColBH 13 54.6 0.69 0.98 1.26 1.79 1.93 2.73 0.82
ColAS 13 43.4 0.79 1.06 1.82 2.45 2.21 2.98 1.08
ColCS 13 39.2 0.73 0.85 1.85 2.17 2.03 2.38 0.94
ColCG 12 8.1 0.41 0.62 5.08 7.60 1.16 1.73 2.60
ColMB 11 22.4 2.77 6.68 12.40 29.85 7.77 18.70 11.91
ColRS 10 20.3 1.15 3.29 5.67 16.19 3.23 9.21 6.37

a
SD = Study Directors laboratory; BH = barley hay; AS = alfalfa silage; CS = corn silage; CG = corn grain; MB = meat and bone meal; RS =
roasted soybeans; Col = collaborators laboratories.
b
2.8 sr.
c
2.8 sR.
d
Ratio of RSDR divided by expected RSD from the Horwitz equation (16).

Two forage samples, BH and AS were selected to represent 14 collaborators had unacceptably high values for CS and 3
feeds that were easy to analyze. Deviations from the results of were unacceptably high for CG. One participant obtained un-
the Study Directors laboratory were used to determine if labo- acceptably low values for CS or CG. The high aNDF results
ratories were following the aNDF procedure exactly. The NDF suggest that starch was not adequately removed from fiber
method is commonly used for feed analysis, and most of the residues. This raised concern that amylase standardization by
participating laboratories were determining NDF routinely. the participants was not adequate. However, from question-
There was concern that these laboratories would use their naire responses there did not seem to be a relationship be-
method rather than the aNDF method that was being evaluated. tween high results and the dose of amylase used by collabora-
For BH, only one participant exceeded the Study Directors tors. If high results were not related to amylase, this suggests
mean +2.8 sR, and both the mean and sR of all collaborators that filtering difficulties caused problems or that participants
agreed closely with those of the Study Director (Table 5). For were not following the aNDF method exactly.
AS, the results of the collaborators were more variable than that Two samples, MB and RS, were known to be extremely
of the Study Director. Two laboratories exceeded the Study Di- difficult to analyze and were included in the familiarization
rectors mean +2.8 sR, and 2 were less than the Study Direc- study to test the limits of the collaborators ability to handle
tors mean 2.8 sR. The sR of all collaborators for AS was difficult samples. The mean results of the collaborators were
nearly triple that of the Study Director, and their HORRAT ap- higher than those determined by the Study Directors labora-
proached 2 (Table 5), suggesting that they were not following tory (Table 5). For MB and RS, 3 participants reported unac-
the aNDF method exactly. Comments on a questionnaire ask- ceptably high values. For RS, 2 collaborators reported unac-
ing about specifics of their technique confirmed that collabora- ceptably low values. The sR of the collaborators was 34 times
tors were deviating from the proposed method, and they were greater than that observed by the Study Director. Some partici-
requested to follow the aNDF method exactly. pants analyzed these materials 5 or more times without obtaining
Two samples, CS and CG, were included in the familiar- acceptable results. Their inability to match the Study Directors
ization set to determine the ability of collaborating laborato- results for CS, CG, MB, and RS convinced most laboratories that
ries to analyze feeds containing starch. The mean values for their method of measuring NDF was inadequate. All collabora-
the collaborators were similar to the Study Directors values, tors were requested to follow the aNDF method exactly when an-
but the collaborators sR was 34 times higher than the Study alyzing the collaborative study samples.
Directors (Table 5). Several laboratories had to analyze the It was evident from both the results of the familiarization
samples numerous times to achieve acceptable results. Some study and comments made by the collaborators that the aNDF
participants only analyzed the samples in duplicate one time method needed to be modified so that laboratories could ana-
whether or not they achieved acceptable results. Thus, 4 of the lyze extremely difficult samples. After the familiarization
1230 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002

study, the aNDF method was modified to improve the analysis Table 6. Doses of 3 amylase stock solutions
of materials that are difficult to filter. Although glass microfiber determined by collaborating laboratories to provide
GF/D mats were used successfully in the Study Directors labo- adequate enzymatic activity for the aNDF method during
ratory, they filtered slowly and were not an acceptable filter aid the familiarization study and actual doses used in the
collaborative study
for some collaborators. Sand had been tried previously by the
Study Director as a filter aid without success when it was mixed Familiarization study Collaborative study
with the sample. There also were concerns that fine sand parti- doses, L doses, L
cles or celite might plug the pores in the fritted disks of Gooch Lab Amylase A Amylase B Amylase C Amylase A Amylase B
crucibles. However, a source of acid-washed silica sand with
particle size distribution between 40 and 200 mesh was found,
0 100 60 120 100
which cannot penetrate and plug the pores of coarse porosity
Gooch crucibles. The sand was stable to ND extraction and 1 75 500 400 200
ashing, and had a negligible effect on blank values. This sand 3 100 200 400 200
was evaluated as a filter aid for the difficult-to-filter samples in 4 300 500 500 300
the familiarization and collaborative studies (MB, RS, CB, and 5 75 125 150 250
DM) and greatly improved filtration and duplicate standard de- 6 200 150 300 150
viations (SDs). The sand was also evaluated with 6 other mate-
7 50 125 200 175
rials that are extremely difficult to filter. In all cases, filtration
was easy and SDs of duplicate analyses were much smaller 8 30 150 100 31
(pooled SD = 0.44 versus 0.77) than was possible without the 9 200 200 200 100
sand filter aid. The aNDF protocol was modified to include sil- 10 400 88 200
ica sand as a filtering aid. Collaborating laboratories were asked 11 88 88 200 88
to reanalyze MB and RS using sand as a filtering aid. With 12 300 200 400 200
sand, their sr decreased from 1.15 to 0.77 for RS and their sR de-
13 300 200 200 250
creased from 6.68 to 2.68 for MB.
14 200 100 150 125
Amylase Standardization Mean 178 202 262 176 170
SD 119 139 124 110 54
Development of the aNDF method was hampered by
changes in the sources of heat-stable -amylase. Sources
would become unavailable or simply be changed by the pro-
vider, making it difficult to specify a source and level of en-
zyme needed for the aNDF method. To eliminate dependence
of the aNDF method on a specific source of amylase and to The mean and median dose of amylase selected by the par-
comply with AOAC policy for Determination of Equivalency ticipants were higher than those selected in the Study Directors
of Products Used in Official Methods, we developed a proce- laboratory. Most collaborators overestimated the dose of amy-
dure for standardizing the heat-stable amylase. Because many lase needed for the aNDF method. This is acceptable because
feed testing laboratories do not have colorimeters, we devised too much amylase is better than too little, although there is a
a simple visual test for standardizing the amylase. In addition, concern that doses >5 times the minimum dose needed may
we found that ND interferes with measurement of starch in cause retrogradation of starch hydrolysis. Except for Labora-
most colorimetric methods, and concluded that a visual test tories 4, 5, and 8, standardization of amylase for the collabora-
would encourage routine evaluation of the amylase. tive study was much more consistent. Laboratories 4 and 8 may
The procedure for standardizing amylase for the aNDF not have restandardized the amylase for the collaborative study
method is based on the reaction of iodine with starch. A proce- samples and simply used the information they collected during
dure was developed that would ensure that raw corn starch the familiarization study. Laboratories 3, 6, 7, and 1012 ob-
(one of the most difficult starches to remove from fiber) would served that amylase C was half the activity of amylase B.
be solubilized and extracted in hot ND solution. It became evi- Several collaborators commented that amylase standard-
dent that the starch-iodine complex that creates a purple color ization was difficult. The procedure was rewritten to clarify
was unstable in ND solution, making visual detection of resid- the steps. There was some indication that participants could
ual starch difficult and time dependent. The original amylase not discriminate between the pink-amber of unhydrolyzed
standardization procedure was evaluated in the Study Direc- starch and the brown of the stock enzyme solutions. Based on
tors laboratory and gave variable results among technicians. results of the familiarization and collaborative studies, we
Additional research indicated that the procedure was sensitive modified the procedure for standardizing amylase for the
to both time and temperature; therefore, it was extensively re- aNDF method to indicate that a control solution of amylase in
written before the collaborative study began. Collaborating blank ND (no grits) should be run with the test doses to improve
laboratories were asked to evaluate 3 amylase stock solutions discrimination between pink-amber (inadequate amylase) and
during the familiarization study to evaluate the adequacy of brown (amylase solution). Although the range in doses of amy-
the amylase standardization procedure (Table 6). lase used was greater than desired, there was no significant rela-
MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1231

tionship in the familiarization study between dose of amylase For aNDF, the HORRAT was <2 for all materials except
and aNDF results across samples and laboratories. DM, RS, and MR (Table 11). Horwitz suggested that a
HORRAT of 2 is a reasonable, if not liberal, criterion for ac-
Evaluation of the aNDF Method cepting a method. The feeds with HORRAT >2 had mean
aNDF <15%. The smaller means for aNDF of these materials
Results of aNDF analyses were calculated 4 different ways greatly inflated the observed RSDR. The Horwitz (16) equa-
to evaluate the effects of blank correction and adjusting fiber tion for calculating the expected RSD was developed by re-
to an organic matter (ash-free) basis (Tables 710). Outliers gressing sR versus the mean concentration across analytes and
were detected by the Cochran and Grubbs tests in AOAC methods. Implicit in this equation is the assumption that limits
spreadsheet software and are shown in Tables 710. Suspect of detection (LODs) decrease and precision increases as the
results were identified as having large residuals (>3*root mean of the analyte decreases. In general, instrumental meth-
mean squared error) in an ANOVA using Model 3 (see Statis- ods for measuring analytes in parts-per-billion and smaller
tical Analyses section) with main effects of Material, Lab, and concentrations (near zero analyte) have very small LODs and
Rep. These latter results were not removed when AOAC sta- very good precision. However, within a method, the relation-
tistics were calculated. To reduce the effect of single suspect ship between precision and mean concentration may differ
results on laboratory ranking and means, the results in bold from the Horwitz equation, especially if the method is
font were removed. An ANOVA using the main-effects gravimetric and attempts to measure analytes from 0 to 100%
model (Model 3) was used to calculate least-squared means concentrations.
for laboratories with all suspect data removed. All outlier data As discussed by Horwitz et al. (12), gravimetric methods
were double checked to verify data entry and to determine if have LODs and precision that are related to weighing error.
collaborators comments could explain the aberrant results. Thus, within the aNDF method, it may be more valid to evalu-
The differences between blind duplicates for grass hay (GH) ate acceptability across materials by comparing sR rather than
and sawdust (SD) reported by Laboratory 11 were not de- RSDR. For aNDF, the sR for DM, RS, and MR are similar to
tected as outliers by ANOVA or AOAC statistical methods. those for other materials, indicating that the method can be
Laboratory 12 had significant or nearly significant Grubb used to analyze feeds with <15% aNDF with precision similar
tests for several materials and reported the highest results for to those with >15% aNDF.
most materials. The laboratory ranking test (14) indicated that Blanks may account for systematic weighing variation
Laboratory 12 was a statistical outlier. Although it reported among runs. The potential magnitude of blank variation within
consistent results for blind duplicates (except for RS), it had a single laboratory is indicated in Tables 3 and 4. Blank varia-
an average bias of +1.99 (1.21) %-units for all materials, and tion among runs in the Study Directors laboratory (2.8 sR) is
questionnaire results indicated that it used medium porosity about 4 mg. The blank variation among collaborating laborato-
Gooch crucibles with ceramic fiber as a filter aid. Neither of ries was greater with a 2.8 sR of 10 mg (Tables 9 and 10). Be-
these modifications was indicated in the aNDF method for cause weights of fiber residues are small, the effect of adjusting
collaborative study. Based on both statistical and methodolog- for systematic weighing variation using blanks should be
ical reasons, the results of Laboratory 12 were not used in the greater for materials with low aNDF. Assuming a 0.5 g sample
statistical evaluation of the aNDF method. and 25% aNDF in the material yields a fiber residue of 125 mg,
Most outliers occurred in 2 laboratories (8 and 9), and in this could result in a range in blank-correction of about 8%. If
each case only one of the blind duplicates was suspect. This the material contains 10% aNDF, the fiber residue would weigh
suggests that the aNDF method and its description may not be only 50 mg and the range in blank-correction could be 20%.
at fault, because most remaining collaborators produced ac- When aNDF was blank-corrected (aNDFbc), the average
ceptable results; even the 2 laboratories with problems gener- sR for materials containing <25% aNDF decreased from 1.30
ated one acceptable result for all materials. Horwitz et al. (12) to 1.01 (Table 11). Thus, blank-correction increases analytical
reported that the reproducibility of empirical methods such as precision for materials with <25% aNDF. For unexplained
fiber is generally much poorer than for analytes such as nitro- reasons the sR for materials containing >25% aNDF increased
gen, which are well defined and have suitable reference stan- from 1.29 to 1.36 due primarily to an increase in sR for
dards. Because the analyte is defined by the method, it is cru- brewers grains (BG). It is difficult to envision how correcting
cial that fiber methods be followed exactly. As Horwitz et for blanks could harm the precision of materials with >25%
al. (12) surmized, laboratories often modify methods to fit lab- aNDF, given that the magnitude of blank-correction diminishes
oratory conditions, attempt to improve the method by elimi- as the fiber residue increases. The HORRAT of aNDFbc was
nating or changing steps or apparatus, or simply ignore details <2 for all materials except BG, DM, MR, and RS (Table 11).
of the method that are thought to be superfluous. Although the The HORRAT increased slightly for BG, but decreased for
familiarization study helped to establish that the aNDF DM, MR, and RS when aNDF was blank-corrected.
method should be followed as described, we suspected that The nutritional definition of fiber, which is the guide for
some laboratories used modifications to accommodate their developing appropriate methods, indicates that fiber is the in-
unique laboratory situation. A questionnaire was used to dis- digestible or slowly digesting fraction of a feed that occupies
cover exactly how each participant measured aNDF and the space in the digestive tract of animals; therefore, indigestible
responses were used to explain atypical results. ash should be included in fiber. However, fiber traditionally
1232 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002
Table 7. Results of collaborating laboratories for amylase-treated neutral detergent fiber (aNDF) expressed on as-received basis
aNDF, % Collaborating laboratory

Feed Duplicate 0 1 3 4 5 6 7 8 9 11 12a 13 14

Alfalfa silage 1 40.51 41.13 41.11 40.08 42.36 38.43 41.49 41.29 39.19 40.92 44.57 39.60 39.84
2 40.26 40.20 40.82 40.88 41.26 40.66 41.43 42.57 40.71 42.20 45.85 41.40 41.20
Brewers grains 1 49.67 46.55 50.19 49.00 52.76 49.25 49.46 48.60b 46.28 51.41 52.25 52.00 48.87
c b
2 47.11 49.30 50.37 48.25 50.64 49.10 44.92 16.18 46.78 47.05 54.11 50.40 47.98
Citrus and beet pulp 1 29.30 30.13 29.69 28.73 31.42 29.06 27.57 29.33 28.45 29.96 31.91 31.20 30.46
2 28.64 30.77 30.63 31.68 30.05 30.87 29.58 29.17 29.67 28.58 32.73 32.00 28.84
Corn grain with cob 1 21.13 21.91 22.79 21.89 21.79 19.60 21.57 22.20 25.50b 22.08 23.72 19.80 21.92
2 22.03 21.60 23.06 21.69 21.81 19.93 21.62 21.47 20.61b 21.12 25.80 21.40 22.10
Corn silage 1 36.43 36.82 38.38 37.08 37.89 36.81 35.94 37.24 35.98 38.88 39.28 37.20 37.10
2 36.05 35.92 38.03 36.42 37.36 35.52 38.66 38.22 35.54 37.23 39.64 37.60 37.42
Corn stalks 1 71.42 72.68 74.30 70.85 74.00 72.05 73.92 74.20b 72.48 74.56 75.03 74.40 73.04
2 70.83 73.40 74.15 71.93 72.83 72.21 73.38 98.79b 73.22 73.05 76.49 74.00 72.26
Dairy mixed feed 1 12.30 12.40 12.90 11.79 14.22 10.87 11.79 13.27 11.14 13.45 16.26 14.60 12.23
2 11.74 12.78 11.62 9.06 12.17 13.18 13.90 14.44 11.76 12.92 15.86 14.80 12.50
b
Grass hay 1 55.66 57.57 58.06 58.21 57.57 57.75 58.79 57.40 48.46 59.65 59.96 59.60 57.56
2 55.41 57.56 58.90 59.48 58.44 58.21 57.62 59.52 56.87b 55.92 62.16 60.40 58.46
d
Milk replacer 1 0.14 0.52 0.62 0.45 0.14 0.38 1.36 0.66 0.00 0.18 1.37 2.40 0.36
2 0.02 0.12 0.28 0.38 0.12 1.83 0.78 3.72d 0.40 0.12 1.65 0.80 0.30
e b
Roasted soybeans 1 12.45 12.75 13.95 8.59 19.13 11.31 14.59 11.91 15.73 15.09 19.11 12.40 13.54
2 11.96 13.63 15.92 11.93e 12.05b 10.82 16.78 11.74 17.83c 14.69 26.00c 12.20 13.24
Sawdust 1 92.01 88.23 90.54 90.49 90.59 91.01 87.93e 87.93 90.61b 88.01 90.19 90.20 90.60
e
2 89.87 89.33 90.44 90.99 90.48 90.98 83.70 89.79 75.10b 93.46 92.15 92.60 91.32
f
Least-squared means 37.95 38.42 39.40 38.33 39.27 37.97 38.88 38.80 37.99 39.12 41.66 39.30 38.69

a
Outlier detected by laboratory ranking.
b
Outlier detected by Cochran test.
c
Outlier detected by analysis of variance to determine least-squared means, but not removed for AOAC statistics.
d
Outlier in Tables 810.
e
Outlier detected by single Grubb test.
f
Least-squared means calculated without data in bold font.
Table 8. Results of collaborating laboratories for amylase-treated neutral detergent fiber organic matter (aNDFom) expressed on as-received basis
aNDF, % Collaborating laboratory

Feed Duplicate 0 1 3 4 5 6 7 8 9 11 12a 13 14

Alfalfa silage 1 38.70 39.37 39.70 39.96 40.48 37.24 39.73 39.55 38.15 39.66 42.02 40.80 37.88
2 38.25 38.45 39.40 39.46 39.48 40.35 39.23 40.62 39.75 40.41 43.04 39.60 39.28
b
Brewers grains 1 48.20 45.30 49.27 48.74 51.64 48.36 48.10 48.04 45.94 50.26 50.72 53.00 49.20
2 45.51 48.05 49.20 49.60 49.36 48.79 43.52c 16.02b 45.30 46.29 52.34 49.40 47.20
Citrus and beet pulp 1 27.20 27.80 27.54 27.68 28.80 27.92 25.49 26.95 27.11 28.40 28.22 29.40 27.42
2 25.95 28.50 28.90 28.53 27.15 29.25 26.74 27.66 26.93 26.98 29.03 28.60 27.12
b
Corn grain with cob 1 20.88 21.55 22.49 22.71 21.57 20.89 21.23 21.45 25.76 21.71 22.76 21.20 21.40
2 21.32 21.25 23.02 22.16 21.63 21.55 20.96 21.25 20.87b 21.04 24.57 21.60 21.78
Corn silage 1 35.76 36.26 37.72 37.44 37.03 36.61 35.30 36.04 35.94 38.04 37.96 37.40 36.20
2 35.12 35.28 37.53 36.74 36.74 36.58 37.02 37.49 35.24 37.07 38.22 37.00 36.76
b
Corn stalks 1 67.18 68.12 71.46 68.62 69.90 68.75 70.76 70.00 69.03 71.37 70.28 72.40 69.52
b
2 67.07 70.31 70.37 69.29 68.88 70.34 67.96 82.74 69.45 70.08 71.82 71.80 68.62

MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1233


Dairy mixed feed 1 11.46 11.98 12.78 12.46 13.52 11.71 12.01 12.78 11.78 13.39 14.63 13.40 12.93
2 10.69 11.99 11.46 11.06 12.01 13.40 12.70 13.50 11.68 12.86 15.96 12.60 12.48
Grass hay 1 53.94 55.24 55.92 57.49 55.22 55.42 56.41 54.83 48.34b 57.24 56.81 57.40 55.48
2 53.48 55.14 57.05 57.94 56.65 57.24 55.40 57.59 55.68b 53.93 59.29 58.80 56.58
b
Milk replacer 1 0.26 0.52 0.78 0.57 0.14 0.34 0.22 0.58 1.55 0.26 0.64 0.40 0.20
b
2 0.02 0.22 0.67 0.73 0.16 0.10 0.44 2.84 0.36 0.53 0.38 0.60 0.20
Roasted soybeans 1 13.01 12.00 14.03 11.80 18.61b 12.64 14.61 11.22 17.04 15.23 17.95 13.60 13.84
2 12.32 13.19 15.48 13.22 11.75b 11.47 15.54 11.02 17.95c 14.77 24.77c 13.00 13.46
d b
Sawdust 1 91.05 87.15 89.63 90.39 89.40 86.26 86.83 86.29 90.19 86.69 90.09 90.40 89.32
d b
2 88.65 88.10 89.67 90.67 89.50 90.42 82.10 88.62 74.41 92.25 90.94 91.40 90.08
e
Least squared means 36.64 37.08 38.37 38.06 37.94 37.53 37.45 37.41 37.30 38.11 39.70 38.80 37.59

a
Outlier detected by laboratory ranking.
b
Outlier detected by Cochran test.
c
Outlier detected by analysis of variance to determine least-squared means, but not removed for AOAC statistics.
d
Outlier detected by single Grubb test.
e
Least-squared means calculated without data in bold font.
1234 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002
Table 9. Results of collaborating laboratories for amylase-treated neutral detergent fiber blank-corrected (aNDFbc) and expressed on as-received basis
aNDF, % Collaborating laboratory

Feed Duplicate 0 1 3 4 5 6 7 8 9 11 12a 13 14

Alfalfa silage 1 40.58 41.11 40.33 40.31 42.34 39.84 41.23 41.25 39.16 40.06 43.25 40.10 39.64
2 40.32 40.18 40.04 41.11 41.24 42.07 41.17 42.53 40.68 41.34 44.53 41.90 41.00
b
Brewers grains 1 49.74 46.53 49.40 49.22 52.74 50.65 49.20 48.44 46.25 50.55 50.93 52.50 48.75
2 47.17 49.28 49.58 48.46 50.62 50.52 44.66c 16.02b 46.56 46.19 52.79 50.90 47.91
Citrus and beet pulp 1 29.37 30.11 28.91 28.96 31.40 30.48 27.31 29.29 28.42 29.09 30.58 31.70 30.26
2 28.70 30.75 29.85 31.90 30.03 32.28 29.32 29.13 29.64 27.71 31.41 32.50 28.77
b
Corn grain with cob 1 21.21 21.89 22.01 22.13 21.77 21.01 21.31 22.16 25.47 21.22 22.39 20.30 21.72
2 22.10 21.59 22.28 21.91 21.79 21.34 21.36 21.43 20.58b 20.26 24.48 21.90 21.90
Corn silage 1 36.50 36.80 37.59 37.31 37.87 38.21 35.68 37.20 35.95 38.01 37.95 37.70 36.90
2 36.12 35.90 37.25 36.65 37.34 36.94 38.40 38.18 35.51 36.36 38.31 38.10 37.22
Corn stalks 1 71.49 72.66 73.52 71.08 73.98 73.46 73.66 74.17b 72.45 73.70 73.71 74.90 72.84
2 70.90 73.38 73.36 72.16 72.81 73.61 73.12 98.63b 73.19 72.19 75.16 74.50 72.06
Dairy mixed feed 1 12.37 12.38 12.12 12.03 14.20 12.29 11.53 13.23 11.11 12.59 14.94 15.10 12.12
2 11.80 12.76 10.84 12.44 12.35 14.59 13.64 14.40 11.73 12.05 14.54 15.30 12.30
Grass hay 1 55.73 57.55 57.28 58.44 57.55 59.16 58.53 57.36 48.43b 58.78 58.64 60.10 57.36
2 55.47 57.54 58.11 59.71 58.42 59.62 57.36 59.48 56.84b 55.05 60.84 60.90 58.26
b
Milk replacer 1 0.21 0.50 0.17 0.23 0.12 1.04 1.62 0.62 0.03 0.69 0.05 1.90 0.16
b
2 0.09 0.10 0.50 0.60 0.10 0.43 0.52 3.68 0.18 0.74 0.33 0.30 0.10
Roasted soybeans 1 12.90 12.73 13.18 12.04 19.11b 12.72 14.33 11.87 15.70 14.23 17.78 12.90 13.34
2 12.62 13.61 15.14 12.14 12.03b 12.22 16.52 11.70 17.80c 13.83 24.68c 12.70 13.17
d e
Sawdust 1 92.08 88.21 89.76 90.72 90.57 92.41 87.67 87.89 90.58 87.14 88.87 90.70 90.40
2 89.94 89.31 89.66 91.22 90.46 92.39 83.44d 89.75 75.07e 92.60 90.83 93.10 91.12
Least-squared meansf 38.06 38.40 38.62 38.65 39.25 39.38 38.62 38.75 37.93 38.25 40.34 39.80 38.51
Blanks 1 0.0007 0.0004 0.0051 0.0007 0.0001 0.0066 0.0022 0.0003 0.0007 0.0034 0.0056 0.0070 0.0012
2 0.0000 0.0002 0.0028 0.0016 0.0001 0.0076 0.0004 0.0001 0.0010 0.0053 0.0077 0.0020 0.0008

a
Outlier detected by laboratory ranking.
b
Outlier detected by Cochran test.
c
Outlier detected by analysis of variance to determine least-squared means, but not removed for AOAC statistics.
d
Outlier detected by single Grubb test.
e
Outlier in Tables 7, 9, and 10.
f
Least-squared means calculated without data in bold font.
Table 10. Results of collaborating laboratories for amylase-treated neutral detergent fiber organic matter blank-corrected (aNDFombc) and expressed on
as-received basis
aNDF, % Collaborating laboratory

Feed Duplicate 0 1 3 4 5 6 7 8 9 11 12a 13 14

Alfalfa silage 1 38.47 39.24 38.97 38.76 40.43 37.67 39.54 39.54 37.50 38.85 41.58 40.60 37.78
2 38.33 38.32 38.67 38.30 39.43 40.78 39.04 40.61 39.11 39.59 42.60 39.40 39.18
Brewers grains 1 47.97 45.17 48.54 47.60 51.59 48.79 47.91 48.00b 45.30 49.44 50.28 52.80 49.08
2 45.59 47.92 48.46 48.50 49.31 49.22 43.33c 15.98b 45.08 45.48 51.91 49.20 47.13
Citrus and beet pulp 1 26.96 27.67 26.81 26.50 28.75 28.36 25.30 26.94 26.46 27.58 27.78 29.20 27.32
2 26.03 28.37 28.17 27.41 27.10 29.69 26.55 27.65 26.28 26.16 28.59 28.40 27.05
Corn grain with cob 1 20.64 21.42 21.76 21.52 21.52 21.32 21.04 21.44 25.11b 20.89 22.32 21.00 21.30
2 21.40 21.12 22.29 21.00 21.58 21.99 20.77 21.24 20.23b 20.23 24.13 21.40 21.68
Corn silage 1 35.53 36.13 36.99 36.27 36.98 37.05 35.11 36.03 35.29 37.23 37.52 37.20 36.10
2 35.20 35.15 36.80 35.55 36.69 37.02 36.83 37.48 34.60 36.25 37.78 36.80 36.66
Corn stalks 1 66.95 67.99 70.73 67.44 69.85 69.19 70.57 69.99b 68.38 70.56 69.84 72.20 69.42
2 67.15 70.18 69.63 68.11 68.83 70.77 67.77 82.71b 68.80 69.27 71.38 71.60 68.52

MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1235


Dairy mixed feed 1 11.23 11.85 12.05 11.27 13.47 12.15 11.82 12.77 11.13 12.58 14.19 13.20 12.82
2 10.77 11.86 10.73 11.27 11.51 13.84 12.51 13.49 11.03 12.04 15.52 12.40 12.38
Grass hay 1 53.71 55.11 55.19 56.32 55.17 55.86 56.23 54.82 47.69b 56.43 56.37 57.20 55.38
2 53.56 55.01 56.31 56.77 56.60 57.68 55.21 57.58 55.03b 53.11 58.85 58.60 56.48
b
Milk replacer 1 0.03 0.39 0.04 0.58 0.09 0.78 0.03 0.57 0.90 0.56 0.20 0.20 0.10
2 0.10 0.09 0.06 0.43 0.11 0.53 0.25 2.83b 0.14 0.28 0.06 0.40 0.10
Roasted soybeans 1 12.72 11.88 13.30 12.02 18.56b 13.08 14.42 11.21 16.40 14.42 17.51 13.40 13.74
2 12.19 13.06 14.75 12.13 11.70b 11.90 15.35 11.01 17.30c 13.96 24.33c 12.80 13.39
d b
Sawdust 1 90.82 87.02 88.90 89.21 89.35 86.70 86.64 86.28 89.54 85.88 89.65 90.20 89.22
2 88.73 87.97 88.94 89.50 89.45 90.86 81.91d 88.61 73.76b 91.44 90.50 91.20 89.98
Least-squared meanse 36.55 36.95 37.64 37.02 37.87 37.97 37.26 37.40 36.70 37.30 39.25 38.61 37.49
Blanks 1 0.0004 0.0009 0.0035 0.0049 0.0002 0.0047 0.0020 0.0000 0.0047 0.0015 0.0021 0.0020 0.0008
2 0.0004 0.0004 0.0039 0.0068 0.0003 0.0003 0.0001 0.0001 0.0018 0.0067 0.0023 0.0000 0.0002

a
Outlier detected by laboratory ranking.
b
Outlier detected by Cochran test.
c
Outlier detected by analysis of variance to determine least-squared means, but not removed for AOAC statistics.
d
Outlier detected by single Grubb test.
e
Least-squared means calculated without data in bold font.
1236 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002
Table 11. Statistical data for aNDF and aNDFbc results
Single Double
Sample ID Outlier n Mean, % sr sR RSDr, % RSDR, % 2.8 sr 2.8 sR HORRAT Cochran Grubb Grubb

aNDF

Alfalfa silage 12 40.81 0.88 0.97 2.15 2.38 2.46 2.71 1.04 0.269 11.4 19.3
Brewers grains 8 11 48.97 1.68 1.98 3.44 4.05 4.72 5.56 1.82 0.330 13.4 29.4
Citrus and beet pulp 12 29.82 1.05 1.14 3.51 3.81 2.93 3.19 1.59 0.331 20.1 28.2
Corn grain with cob 9 11 21.57 0.49 0.87 2.25 4.04 1.36 2.44 1.60 0.493 29.8 49.3
Corn silage 12 37.07 0.79 0.97 2.13 2.63 2.21 2.73 1.13 0.491 10.2 16.7
Corn stalks 8 11 72.95 0.58 1.13 0.79 1.55 1.62 3.17 0.74 0.312 13.7 32.2
Dairy mixed feed 12 12.58 1.04 1.33 8.23 10.59 2.90 3.73 3.88 0.291 16.8 36.6
Grass hay 9 11 58.08 1.06 1.31 1.82 2.25 2.96 3.65 1.04 0.566 35.3 63.3
a
Milk replacer 8 11 0.13 0.69 0.81 547.15 644.34 1.92 2.27 0.440 31.0 66.1
Roasted soybeans 5 11 13.32 1.08 2.18 8.13 16.36 3.03 6.10 6.04 0.433 12.8 22.0
Sawdust 7 and 9 10 90.44 1.51 1.51 1.67 1.67 4.22 4.22 0.82 0.653 19.8 67.4
aNDFbc

Alfalfa silage 12 40.81 0.88 0.88 2.15 2.15 2.46 2.46 0.94 0.268 13.4 33.8
Brewers grains 8 11 48.19 1.85 2.24 3.83 4.64 5.17 6.26 2.08 0.280 11.9 23.5
Citrus and beet pulp 12 29.83 1.04 1.39 3.48 4.66 2.91 3.89 1.94 0.336 16.6 29.0
Corn grain with cob 9 11 21.57 0.48 0.55 2.25 2.55 1.36 1.54 1.01 0.496 19.3 29.1
Corn silage 12 37.07 0.79 0.86 2.13 2.31 2.21 2.40 1.00 0.493 20.1 30.5
Corn stalks 8 11 72.96 0.58 1.04 0.79 1.43 1.62 2.91 0.68 0.311 16.5 35.9
Dairy mixed feed 12 12.72 0.86 1.23 6.73 9.64 2.40 3.43 3.53 0.301 28.7 38.6
Grass hay 9 11 58.08 1.06 1.49 1.82 2.57 2.96 4.18 1.18 0.567 18.7 40.7
a
Milk replacer 8 11 0.13 0.68 0.68 519.84 519.84 1.92 1.92 0.443 22.9 38.5
Roasted soybeans 5 11 13.52 0.81 1.59 5.99 11.80 2.27 4.46 4.36 0.332 26.9 46.1
Sawdust 7 and 9 10 90.47 1.51 1.61 1.67 1.78 4.23 4.50 0.87 0.653 12.4 24.1

a
= HORRAT could not be calculated because the mean was a negative number.
Table 12. Statistical data for aNDFom and aNDFombc results
Single Double
Sample ID Outlier n Mean, % sr sR RSDr, % RSDR, % 2.8 sr 2.8 sR HORRAT Cochran Grubb Grubb

aNDFom

Alfalfa silage 12 39.40 0.91 0.91 2.32 2.32 2.56 2.56 1.01 0.484 7.0 15.9
Brewers grains 8 11 48.19 1.85 2.24 3.83 4.64 5.17 6.26 2.08 0.280 11.9 23.5
Citrus and beet pulp 12 27.67 0.76 1.00 2.76 3.61 2.14 2.79 1.49 0.195 14.7 26.2
Corn grain with cob 9 11 21.57 0.31 0.58 1.46 2.69 0.88 1.62 1.07 0.204 28.1 65.7
Corn silage 12 36.60 0.61 0.85 1.67 2.33 1.71 2.38 1.00 0.329 8.9 19.8
Corn stalks 8 11 69.60 0.97 1.46 1.40 2.09 2.73 4.08 0.99 0.376 19.5 44.4
Dairy mixed feed 12 12.36 0.70 0.81 5.63 6.54 1.95 2.26 2.39 0.246 18.5 24.3
Grass hay 9 11 56.11 1.17 1.43 2.08 2.55 3.26 4.01 1.17 0.368 22.7 39.9
Milk replacer 8 11 0.42 0.29 0.33 68.35 78.87 0.81 0.93 17.31 0.777 21.9 34.4
Roasted soybeans 5 11 13.66 0.67 1.86 4.90 13.61 1.87 5.20 5.04 0.214 25.6 34.9
Sawdust 7 and 9 10 89.30 1.76 1.76 1.97 1.97 4.93 4.93 0.97 0.498 12.5 33.3
aNDFombc

MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1237


Alfalfa silage 12 39.09 0.91 0.91 2.32 2.32 2.54 2.54 1.01 0.489 8.2 20.1
Brewers grains 8 11 47.88 1.82 2.24 3.80 4.68 5.10 6.27 2.09 0.288 13.0 28.6
Citrus and beet pulp 12 27.36 0.75 1.08 2.73 3.95 2.09 3.03 1.63 0.204 12.6 29.0
Corn grain with cob 9 11 21.30 0.34 0.46 1.58 2.16 0.94 1.29 0.86 0.229 17.6 38.3
Corn silage 12 36.29 0.60 0.82 1.66 2.27 1.68 2.30 0.97 0.339 16.1 28.4
Corn stalks 8 11 69.27 0.98 1.46 1.41 2.10 2.73 4.07 0.99 0.375 22.8 41.1
Dairy mixed feed 12 12.09 0.67 0.89 5.55 7.38 1.88 2.50 2.69 0.358 6.6 15.8
Grass hay 9 11 55.83 1.16 1.38 2.08 2.48 3.25 3.87 1.14 0.370 20.4 42.6
Milk replacer 8 11 0.11 0.21 0.37 191.63 347.64 0.58 1.05 62.15 0.622 14.8 35.4
Roasted soybeans 5 11 13.38 0.59 1.63 4.39 12.17 1.65 4.56 4.49 0.276 27.9 38.6
Sawdust 7 and 9 10 89.01 1.74 1.74 1.96 1.96 4.87 4.87 0.96 0.510 14.6 34.1
1238 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002

materials contain <25% aNDF, and this is the recommended


method for these feeds and for regulatory laboratories.
Occasionally ash values for fiber residues were negative.
When this occurred, some collaborating laboratories treated
this result as an error, ignored the ash correction, and set
aNDFom = aNDF. Correcting for blanks reduced the incidence
of negative fiber ash results, but did not eliminate them. A nega-
tive ash means that crucible weight after ashing was less than
the initial weight of the crucible and indicates that crucibles lost
weight during the aNDF procedure. The most accurate estimate
of the weight of fiber residues is obtained by using the crucible
weight after ashing, because the initial crucible weight is an
overestimate caused by some contaminant in the crucible that
Figure 1. Change in standard deviation of was lost during the aNDF method, i.e., aNDFom calculated us-
reproducibility, s(R), due to ashing of fiber residues. ing a negative ash is a more accurate estimate of fiber than is
aNDF. The aNDF method submitted for First Action was modi-
fied to indicate that aNDFom results must be calculated using
ashed crucible weight when it is less than the initial crucible
has been defined as organic matter, and in plant materials it
weight (negative ash) because in this circumstance the aNDF
has been equated with plant cell walls, which contain little ash.
result is an underestimate of the correct value.
Recently, nutritionists have derived an estimate of the easily
Sokal and Rohlf (18) suggest that the number of decimal
digested or rapidly fermentable carbohydrates by difference.
places for reporting results should be based on the magnitude
Nonfiber carbohydrates (NFC) or neutral detergent soluble
of the standard error using the following approach: divide the
carbohydrates (NDSC) are calculated as:
standard error by 3 and use the decimal place of the first non-
zero digit to determine the significant digits appropriate for re-
Dry matter crude protein crude fat ash aNDF
porting results. Based on their rule of thumb, it is recom-
If some ash is included in aNDF this results in a double sub- mended that aNDF results be reported to the nearest 0.1%.
traction of the ash in fiber. Thus, there are nutritional benefits Irrespective of the basis for calculating results (aNDF,
to expressing aNDF as ash-free organic matter (aNDFom). aNDFbc, aNDFom, or aNDFombc) it is evident that the RSD in-
creases as the mean concentration of aNDF decreases. The data in
Analytically, ashing fiber residues and expressing results
Tables 11 and 12 indicate that sR does not approach zero as mean
as aNDFom improves the reproducibility of results for materi-
concentration of aNDF approaches zero; rather it approaches some
als with <50% aNDF (Tables 11 and 12). However, it de-
minimum value that can exceed the mean value. This has implica-
creases analytical reproducibility slightly for materials with
>50% aNDF. The consistent inverse relationship between the
change in sR due to ashing of fiber residues and mean concen-
tration of aNDFom is unexplained (Figure 1). A similar trend 2.50
BG
was observed when results were blank-corrected. It was not
2.00
related to the amount of ash in the fiber. In general, expressing Horwitz
Standard Deviation

RS
fiber as aNDFom decreased the HORRAT. Ashing fiber resi- 1.50 Horwitz
dues requires an additional step in the procedure, which incurs s - Col
s r - Col
s R - Col
R s r - Col
extra time and expense. For routine fiber analysis of forages, 1.00 s R - SD
s R - SD
measuring aNDFom may not be cost effective.
The average sR for all materials was 1.29, 1.24, 1.20, and 0.50

1.18 for aNDF, aNDFbc, aNDFom, and aNDFombc, respec-


0.00
tively. Although adjusting results for both ash and blanks im- 0 20 40 60 80 100
proves precision, the small differences among results do not aNDFombc (%)
indicate a strong preference for one method of expressing the
results over another. Thus, it is suggested that the method for Figure 2. Relationship of standard deviation of
expressing the data be left to the discretion of the laboratory reproducibility calculated by the equation of Horwitz
and calculations for all 4 results are given in the aNDF (16) and the standard deviations of repeatability (sr) or
reproductibility (sR) obtained in the collaborative study
method. The laboratory must state clearly which method for (Col) or homogeneity test by the Study Director (SD) to
calculating results was used. The exception to this general rule the mean amylase-treated neutral detergent fiber or-
of allowing analysts to choose the method of expressing ganic matter, blank-corrected (aNDFombc) for each
aNDF results is for feeds containing <25% aNDF. The aver- material. Data points for roasted soybeans (RS) and
age sR for these materials is 1.30, 1.01, 0.90, and 0.84 for brewers grains (BG) were omitted from the sR Col re-
gression line. Intercepts of lines were 0.31, 0.45, and
aNDF, aNDFbc, aNDFom, and aNDFombc, respectively. 0.39 for sr Col, sR Col, and sR SD, respectively.
There is a clear advantage to ash- and blank-correction when
MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002 1239

tions on the minimum level of applicability for the aNDF method. order of analysis, e.g., preceding an analysis with an easy- or
Horwitz et al. (12) concluded that the inherent variability in a difficult-to-filter sample, did not affect results.
gravimetric method in which the analyte is empirically defined In the first blind duplicate, materials were labeled as con-
limits its ability to detect low levels of concentration. taining >10% fat when appropriate. In the second blind dupli-
Figure 2 illustrates that the minimum sR for aNDFombc is cate, the materials were not identified as containing >10% fat to
near 0.4. Regressing the standard deviation versus the mean determine if laboratories could detect unidentified samples that
concentration permits estimation of the repeatability and were high in fat. Only 3 of the 12 collaborators identified some
reproducibility of aNDFombc at zero concentration. Using the of the materials in the second blind replicate that were >10%
data from Table 4, the intercept for reproducibility in the fat, and no laboratory detected all 3 materials. This may explain
Study Directors laboratory was 0.39. Because the data in Ta- why results for RS and DM were less reproducible than those
ble 4 were generated in one laboratory and assume there were for other materials. In most instances, the first blind duplicate
no differences among sample sets, the sR from Table 4 is com- identified as >10% fat was pre-extracted and the second un-
parable to the within-laboratory repeatability from data in Ta- identified blind duplicate was not. Statistical analysis of the re-
ble 12. For the collaborative study, the intercept for repeatabil- sults did not detect a significant difference between samples
ity within laboratories (sr) was 0.31 and the intercept for that were pre-extracted and those that were not; however, the
reproducibility among laboratories (sR) was 0.45 (results for reproducibility for these materials was large, making it difficult
BG and RS were not included in the regression). This suggests to detect differences. Results in the Study Directors laboratory
when the true result is zero concentration, one time in 20 labo- clearly demonstrate that pre-extraction of materials containing
>10% fat improved repeatability of results and lowered aNDF
ratories would observe values <1.3 or >1.3% aNDFombc
values. The inability of collaborators to detect unidentified
(1.3% = 2.8 .45). It is recommended that values <1.5%
samples containing >10% fat and the decreased reproducibility
aNDF be reported as <1.5%.
of these materials suggest that perhaps regulatory laboratories
Laboratories were provided with 2 different amylase should pre-extract all materials before aNDF determination un-
sources and requested to determine the standard dose for their less it is known that they contain <10% fat.
source to use in the collaborative study. An ANOVA was per- An ANOVA using Equation 1 without Block indicated a
formed on the fiber results for all materials and only materials significant Lab*Material interaction when the results for all
containing significant starch (CG and CS). There was no dif- materials were tested. This indicated that the aNDF method
ference in aNDF, aNDFbc, aNDFom, or aNDFombc among was not consistent for all laboratories across all materials.
amylase sources or doses for all materials or for those contain- When the results for high-fiber materials and forages (SD,
ing starch. This suggests that the amylase standardization pro- corn stalks, GH, AS, CS) were combined and tested, the
cedure is adequate for the aNDF method. As indicated in Ta- Lab*Material interaction was not significant. Likewise when
ble 6, most collaborators used doses of amylase higher than concentrated materials containing <10% fat (BG, CB, CG)
those selected by the Study Directors laboratory, which may were tested, the Lab*Material interaction was not significant.
explain why removal of starch from fiber was not a problem. These statistical analyses indicate that collaborators could ob-
Laboratory 8 used a smaller amylase dose than that selected tain consistent results for forages and concentrated feeds with
by the Study Directors laboratory, but obtained acceptable fi- <10% fat. However, Lab*Material interaction was significant
ber values for CG and CS. for concentrated feeds with >10% fat (DM, RS, MR), which
Questionnaire responses were used to evaluate modifica- suggests that collaborators had difficulty obtaining consistent
tions or variations of the aNDF method that were used by col- results for these feeds and confirms the variability associated
laborating laboratories. Some variations were made by individ- with high sR for these feeds. This inconsistency and variability
ual collaborators and others were related to modifications may be related to the variable pre-extraction of these feeds
allowed in the aNDF method. An ANOVA indicated that within- and among-laboratories. Results were pooled (13)
refluxing in crucibles (Fibertec) was not significantly different within categories of feed materials to obtain precision perfor-
from refluxing in beakers and filtering in crucibles. Collaborators mance parameters for the aNDF method (Table 2002.04).
using the Fibertec apparatus had an average bias for the 11 mate-
rials of +0.60% for aNDFbc, aNDFom, and aNDFombc, which Collaborators Comments
was significant but within the sR of the method. When the aver- Several collaborators had been measuring NDF prior to the
age results for each material determined with the Fibertec appa- collaborative study and offered suggestions about their modifi-
ratus was regressed versus the average results with traditional cations of the method. They suggested that extra crucibles be
refluxing in beakers, the intercepts were not different from zero dried and tare weights recorded. When a difficult-to-filter test
and the slopes were not different from 1 for aNDF, aNDFbc, sample was encountered, the residue could be split between
aNDFom, and aNDFombc. This suggests that the Fibertec appa- 2 crucibles so that the analysis could be completed without a re-
ratus is a suitable alternative for measuring aNDF. run. Although this suggestion allowed results to be generated, it
Statistical analysis also indicated that the first replicate in increased the expected variation in results due to additional er-
which materials were ordered from least to most difficult to ror associated with weighing 2 crucibles to obtain results.
filter was not different from the second replicate in which ma- It was suggested that the initial dose of amylase be injected
terials were analyzed in random order. This suggests that the before placing the beaker on the heating unit instead of waiting
1240 MERTENS: JOURNAL OF AOAC INTERNATIONAL VOL. 85, NO. 6, 2002

until the solution is boiling (ca 5 min after placing beaker on heat- Stanley Kobata and Jack Carmany, California Department
ing unit). However, this is not recommended for 2 reasons: of Food and Agriculture, Sacramento, CA
(1) Initial heating of the test sample results in partial David G. Main and Michael S. Allen, Michigan State Uni-
gelatinization of starch and improves hydrolysis by amylase; and versity, East Lansing, MI
(2) heating the solution tends to deactivate any contaminating Xang Moua and Michael Wolf, JL Analytical Services,
fibrolytic enzymes that may be present in crude amylase extracts. Modesto, CA
Some laboratories did not like to use stirring rods to mix James B. Robertson, Cornell University, Ithaca, NY
residues in the crucibles, or used glass instead of Teflon stir- Jim Williams and David Jeffress, Missouri Department of
ring rods. Stirring the residues is important to insure contact Agriculture, Jeffferson City, MO
between fiber particles and washing solutions. This is espe-
cially important for adequate extraction with acetone at the References
end of the method. If fiber residues are damp when acetone is
added the first time, a fine stream of acetone from a squirt bot- (1) Van Soest, P.J., & Wine, R.H. (1967) J. Assoc. Off. Anal.
tle can be used to mix residues and obtain adequate extraction. Chem. 50, 5055
Teflon rods are recommended because particles stick to them (2) Robertson, J.B., & Van Soest, P.J. (1981) in The Analysis of
less readily than to glass rods. Other comments by collabora- Dietary Fiber in Food, W.P.T. James & O. Theander (Eds),
tors were used to improve the method description. Marcel Dekker, Inc., New York, NY, pp 123158
(3) Van Soest, P.J., & Robertson, J.B. (1987) Analysis of For-
Recommendations ages and Fibrous Foods: A Laboratory Manual, Cornell
University, Ithaca, NY
Based on the results of the collaborative study, the Study (4) Jeraci, J.L., & Van Soest, P.J. (1990) in New Developments
Director recommends that Gravimetric Determination of Am- in Dietary Fiber, I. Furda & C.J. Brine (Eds), Plenum Press,
ylase-Treated Neutral Detergent Fiber in Feeds with NY, pp 245265
Refluxing in Beakers or Crucibles be adopted as Official First (5) Van Soest, P.J., Robertson, J.B., & Lewis, B.A. (1991) J.
Action. It is recommended that alternative methods of report- Dairy Sci. 74, 35833597
ing results be allowed, except that materials with <25% aNDF (6) Mertens, D.R. (1988) Recent Modifications to the Neutral
must be blank-corrected, that materials containing >10% fat Detergent Fiber Procedure, U.S. Dairy Forage Research
must be pre-extracted with a suitable solvent before aNDF de- Center Research Summaries, Madison, WI, pp 9798
termination, and that ash-corrected values be reported when (7) Hintz, R.W., Mertens, D.R., & Albrecht, K.A. (1996) J.
negative ash weights are observed. AOAC Int. 79, 1622
(8) McQueen, R.E., & Nicholson, J.W. (1979) J. Assoc. Off.
Acknowledgments Anal. Chem. 62, 676680
(9) Goering, H.K., & Van Soest, P.J. (1970) Forage Fiber Anal-
I thank Diane Amundson, Leslie Sims, and Rodney Hintz ysis (Apparatus, Reagents, Procedures, and Some
(U.S. Dairy Forage Research Center) for assistance in devel- Applications), USDA Agricultural Research Service, Hand-
oping the aNDF method. I also thank Diane Amundson for as- book No. 379, Washington, DC
sistance with preparing and distributing materials for the study (10) Van Soest, P.J. (1965) J. Assoc. Off. Anal. Chem. 48, 785789
and compiling the results for analysis. The USDA Agricul- (11) Mertens, D.R. (1992) Critical Conditions in Determining De-
tural Research Service provided funding for the materials used tergent Fiber, Proceedings of NFTA Forage Analysis
in the study. I also thank the following collaborators for their Workshop, Denver, CO, pp ClC8
participation in the study: (12) Horwitz, W., Albert, R., & Deutsch, M.J. (1990) J. Assoc.
Jane Clegg and Kenneth Frank, University of Nebraska, Off. Anal. Chem. 73, 661680
Soil and Plant Analytical Laboratory, Lincoln, NE (13) Wernimont, G.T. (1990) Use of Statistics to Develop and
Michelle M. Drouches and Dave Taysom, Dairyland Labo- Evaluate Analytical Methods, W. Spendley (Ed.), AOAC,
ratories, Inc., Arcadia, WI Arlington, VA
Michelle Garkie and Bruce Paul, MoorMan Feeds, Inc., (14) Youden, W.J., & Steiner, E.H. (1975) Statistical Manual of
Quincy, IL the Association of Official Analytical Chemists, AOAC,
Bryan Gildemeister and Nancy Thiex, South Dakota State Arlington, VA
University, Brookings, SD (15) Youden, W.J. (1967) Statistical Techniques for Collaborative
Darryl K. Jones and Darrell Gambin, Ralston Analytical Tests, AOAC, Washington, DC
Laboratory, St. Louis, MO (16) Horwitz, W. (1982) Anal. Chem. 54, 67A76A
Deborah Kaplan and Ann Davidowicz, Blue Seal Feeds, (17) Mertens, D.R., Hintz, R.W., Sims, L.S., & Cardoza, R.D.
Inc., Londonderry, NH (1991) Crop Sci. 31, 13611363
Gyoung-No Kim and Cheon-Seok Jeon, JE-IL Feed Co. (18) Sokal, R.R., & Rohlf, F.J. (1981) Biometry, 2nd Ed., W.H.
Chemical Laboratory, Taejeon, Korea Freeman and Co., New York, NY, Ch. 7, p. 151

Potrebbero piacerti anche