Assessment of in situ techniques to determine indigestible components in the feed and feces of cattle receiving supplemental condensed tannins

Abstract

Reliable assessments of indigestible dietary components are required when using internal markers to estimate diet digestibility and determine the potentially digestible portion of the fiber. The lack of a standardized methodology and understanding of how antinutritional factors influence indigestible residues can result in erroneous estimates with inconsistent variation across trials and among studies. Previous studies have detailed suitable bag porosity and sample size (SS) with incubation length (IL) varying from 96 to 504 h, with many assuming that 288-h IL yields truly indigestible components. Recent studies have primarily investigated the variation that exists among feedstuffs, but most have failed to account for possible effects of secondary compounds. Using 2 similar concentrate diets, one of which contained supplemental condensed tannins (CT), we investigated the effect of bag type (BT; 10 and 25 μm), SS (20 and 40 mg/cm²), and IL (288 and 576 h) on in situ indigestible DM (iDM) and neutral detergent fiber (iNDF) residues of feed and feces, and resultant DM and NDF digestibilities. There were no 3-way interactions (P > 0.05), but 2-way interactions were present for iDM and iNDF residues with BT × SS influencing the control (no CT) ration (P < 0.01), SS × IL impacting feed containing CT (P < 0.01), and BT × IL affecting both feedstuffs (P ≤ 0.01). For the control diet, only BT × SS affected DM and NDF digestibilities. Whereas the CT diet did not demonstrate any significant interactions for digestibilities. Values of iDM were largely influenced by contamination that varied greatly based on intrinsic factors associated with the bag and incubation duration. The presence of CT influenced iDM and iNDF to varying degrees due to possible trapping of CT–substrate complexes. For the control diet, the use of 25-μm bags resulted in lower fecal recoveries relative to the 10 μm (P < 0.01). However, there appears to be a dynamic relationship among BT, SS, and IL within respective diets and sample types that can affect indigestible components and resultant digestibility estimates. Based on simulations from these data, the sample size required to attain 90% power when utilizing 2 incubation animals exceeds the triplicate and quadruplicate replications commonly utilized. Further emphasizing the necessity for a more complete understanding of incubation dynamics to design biologically and statistically valid investigations.

Introduction

The proportion of NDF within a feedstuff greatly influences animal performance by affecting organic matter digestibility and total intake, as well as serving as an important source of ME (Harper and McNeill, 2015). However, NDF content is a nutritionally complicated concept because it is a heterogeneous entity that can be fractionated into potentially digestible (pdNDF) and indigestible (iNDF) fractions (Allen and Mertens, 1988). Though not chemically defined, iNDF describes the innate properties of the cell wall and serves as an ideal nutritional entity because its digestibility is zero (Mertens, 1993; Van Soest, 1994), making iNDF integral within nutritional models (Tedeschi and Fox, 2018) and enabling its use as an internal marker (Sampaio et al., 2011; Palmonari et al., 2016). This requires that only the truly indigestible fraction remains following incubation. However, antinutritional factors can potentially confound estimates by altering the degradation rate or being misperceived as indigestible feed components (Van Soest, 2015). Formation of insoluble complexes makes it impossible to determine the true digestibility of the feed particle, resulting in incorrect ration formulation (Tedeschi and Fox, 2018).

The accuracy and precision of iNDF estimates are dependent upon the incubation technique utilized, with bag type (BT) and incubation length (IL) serving as potential sources of error (Nocek, 1988; van der Koelen et al., 1992). Currently, the suggested methodology is to use 288 h incubations and bags of 6- to 12-μm porosity (Krizsan and Huhtanen, 2013; Krizsan et al., 2015). The use of commercially available F57 bags (25 μm; ANKOM Technology, Macedon, NY) for determination of indigestible components has demonstrated acceptable recoveries (Casali et al., 2009; Valente et al., 2011), but particle loss due to bag porosity is a potential source of error when incubating heterogeneous particle sizes for a long period (Huhtanen et al., 1994). A recent commercial bag, F58 (10 μm; ANKOM Technology), provides the potential to reduce bag variability and improve estimations of indigestible components when using heterogeneous particle sizes, but it has not been thoroughly tested under different feeding conditions. The objective of this study was to evaluate the influence of BT, sample size (SS), and IL upon the precision of indigestible DM (iDM) and iNDF estimates, their subsequent digestibilities, and sources of error from feed and feces of animals offered a total mixed ration with and without the addition of condensed tannins (CT), a secondary compound commonly found in certain legumes.

Materials and Methods

The animals used in this experiment were registered and cared for according to guidelines approved by the Institutional Animal Care and Use Committee (AUP’s 2016-0362 and 2017-0306) of Texas A&M University.

Sample Collections

Animals used in this study underwent an indirect calorimetry assay, using respiration chambers. Feed and fecal samples utilized for indigestible component analyses were collected from 6 English crossbred steers (320 ± 21 kg) during a respirometry trial performed in the fall of 2016. Ingredient constitution of the total mixed rations and chemical composition of dietary treatments and corresponding feces used for indigestible component analyses, basal diet (i.e., control) and basal diet with the addition of 3% quebracho (Schinopsis balansae) CT extract, are listed in Table 1.

Table 1.

Ingredient and chemical composition of diets and feces utilized for indigestible content determinations¹

Items2	Basal diet		Basal diet + condensed tannin
	Feed	Feces	Feed	Feces
Ingredient composition, % DM
Dried distillers’ grains	39.38	—	39.38	—
Cracked corn	39.38	—	37.13	—
Bermudagrass hay	13.77	—	12.15	—
Molasses	4.48	—	5.42	—
Mineral	1	—	0.95	—
Urea	0.42	—	0.44	—
Quebracho extract	—	—	3	—
Limestone	1.48	—	1.44	—
Vitamin E	0.09	—	0.09	—
Chemical composition
DM, %	90.08	21.60	90.03	22.73
CP, % DM	20.18	18.66	19.13	24.26
ADF, % DM	10.30	20.71	10.65	23.02
NDF, % DM	20.00	40.78	20.53	38.61
NFC3, % DM	53.33	29.26	53.85	27.97
Ash, % DM	6.52	11.28	6.58	9.17

¹Diets used in the digestion study from which feed and feces were collected for the determination of indigestible components.

²Items are feed ingredients and chemical composition of diets evaluated by Cumberland Valley Analytical Services (Waynesboro, PA).

³NFC, nonfiber carbohydrate.

Upon removal from respiration chambers, a single spot sample of fresh feces was collected directly from the anus, total collection of excreta was not performed, and a subsample of total orts were compiled and stored at −20 °C. Representative samples of each diet were made from 4 samplings during the last 8 d of each period. All samples were dried at 55 °C for 72 h using a forced-air oven and ground to pass through a 2.0-mm screen (Wiley mill, Thomas Scientific, Swedesboro, NJ). A subset of each representative sample was shipped to Cumberland Valley Analytical Services (Waynesboro, PA) for chemical analysis of DM (Goering and Van Soest, 1970), NDF (Van Soest et al., 1991), ADF (Method# 973.18; AOAC, 2012), CP (Method# 990.03; AOAC, 2012; Leco FP-528 Nitrogen Combustion Analyzer, Leco Corporation, St. Joseph, MO), soluble CP (Krishnamoorthy et al., 1982), a complete mineral panel (Method# 985.01; AOAC, 2012; Perkin Elmer 5300 DV ICP, Perkin Elmer, Shelton, CT), and calculation of nonfiber carbohydrates (NFC).

Experimental Design

A complete randomized design using a 2 × 2 × 2 factorial arrangement was used to investigate the effect of BT, SS, and IL upon iDM, iNDF, and subsequent digestibility estimates for basal and CT diets. Diet comparisons were not performed. The determination of indigestible components was performed using F57 or F58 bags (ANKOM Technology, Macedon, NY), 25- and 10-μm porosity respectively, filled with an SS-to-surface area ratio of 20 or 40 mg DM/cm² (20 and 40 mg), and incubated for 288 or 576 h in the rumen of 4 ruminally cannulated English-cross steers (603 ± 15 kg). The F57, 20 mg DM/cm², and 288-h incubation were assumed to be the standard methodology within this study, based on suggested methods (Casali et al., 2009; Volden, 2011; Valente et al., 2011; Krizsan and Huhtanen, 2013; Krizsan et al., 2015).

Within each diet (control and CT), feed and fecal samples for each animal (n = 6) were replicated 16 times for each incubation treatment combination. In total, the number of sample bags per diet exceeded 1,500. Sample bags were placed within 36 × 42 cm polyester bags, 4 polyester incubation bags per cannulated animal. The total weight of all incubation bags represented 0.11% of a cannulated animals’ BW and 12.8% of expected DMI, but animals did not demonstrate any negative responses to the bags and animals maintained BW for the duration of the incubation period. Future studies should recognize that incubating a large number of bags can reduce DMI. Therefore, investigators must ensure that forage and supplement quality enable the animal to meet production requirements with decreased intake. In this study, cannulated animals were maintained in a dry-lot with bermudagrass (Cynodon dactylon) hay (CP = 8.5%, NDF = 77.8%, ADF = 45.2%, NFC = 8% DM) provided ad libitum and supplemented every other day with 250 g of dried distillers’ grains. During preliminary analysis, it was noted that F58 bags sequestered fibrous particles on the outside of the bags; therefore, polyester incubation bags of 50-μm porosity were utilized rather than commercial laundry bags. A preliminary comparison of 50-μm incubation bags with commercial laundry bags was performed to ensure bag porosity did not affect iDM and iNDF residues when using the standard methodology (iDM 4.47 vs. 4.53%, P = 0.57; iNDF 1.51 vs. 1.46%, P = 0.20 for commercial vs. polyester incubation bags, respectively). Timed placement methodology was utilized to ease bag removal and reduce labor. Upon removal from the rumen, all bags were immediately quenched in ice water followed by rinsing in a household washing machine using cold water and the rinsing option of the delicate wash cycle (Krizsan and Huhtanen, 2013) until the water remained clear (45 min).

All bags were dried at 55° C for 72 h using a forced-air oven, placed into desiccators and weighed to obtain iDM. Determination of iNDF was performed using the ANKOM 200 fiber analyzer (ANKOM Technology, Macedon, NY) and washed according to Van Soest et al. (1991) using a detergent-to-sample ratio of 100 mL/g DM and the addition of 4 mL of heat-stable amylase with sodium sulfite being omitted due to its potential for altering residue recovery (Van Soest, 1994, 2015). Bags were then rinsed in hot water followed by soaking in acetone before drying at 55 °C for 48 h. Values of iDM and iNDF were ash inclusive (Sampaio et al., 2011; Valente et al., 2011; Velásquez et al., 2018) due to the number of bags not being conducive to the ashing procedure. However, blank bag corrections were performed for each BT × IL combination. Indigestible DM and NDF were calculated as a proportion of DM using the following equations:

$${}^{\begin{array}{ccc}\text{iDm},\%& =& \frac{\text{W}3-(W1\times \text{C1})}{\text{W2}}\end{array}\times 100}$$ (1)

$${}^{\begin{array}{ccc}\text{iNDF},\%DM& =& \frac{\text{W}4-(W1\times \text{C2})}{\text{W2}}\end{array}\times 100}$$ (2)

where W1 is the initial bag weight (g), W2 is the initial SS (g), W3 is the dried weight of bag containing sample residue following water rinse (g), W4 is the dried weight of bag containing sample residue following washing with ND (g), C1 is the blank bag correction factor following the initial water rinse (running average of final dry bag weight following cold water rinse divided by the initial bag weight), and C2 is the blank bag correction factor following washing with ND (running average of final dry bag weight following ND wash divided by the initial bag weight).

The digestibility of DM (DMD) and NDF (NDFD) were calculated using marker concentrations according to the equations:

$$\begin{array}{ccc}\text{DMD,\%}& =& 100-\left[100\times \left[\frac{M_I}{M_E}\right]\right]\end{array}$$ (3)

$$\begin{array}{ccc}NDFD,\%& =& 100-\left[100\times \left[\frac{M_I}{M_E}\times \frac{{\text{NDF}}_E}{{\text{NDF}}_I}\right]\right]\end{array}$$ (4)

where M_I is the marker concentration of diet consumed (%), M_E is the marker concentration of the fecal sample (%), NDF_I is the NDF concentration of diet consumed (%), and NDF_E is the NDF concentration of the fecal sample (%).

Statistical Analyses

All statistical procedures were performed using SAS software (SAS Institute Inc., Cary, NC). Residuals of indigestible fractions and calculated digestibilities were checked for normality and outliers. Outliers were removed if their distance from the upper or lower quartile exceeded 3 times the interquartile range (Tukey, 1977). Indigestible fractions were evaluated by diet (basal and CT) and sample type (feed and feces) in accordance with a 2 × 2 × 2 factorial arrangement using PROC GLIMMIX with incubation animal (cannulated animal) serving as a random factor in the statistical model. Differences were considered significant at P ≤ 0.05. Digestibility estimations were assessed by diet utilizing the previous model with the donor animal (animal sample was collected from) and collection period (period samples were collected in the feeding trial) as random factors. Statistical model selections were based on corrected Akaike’s information criterion with variance partitioning of random factors for all models being performed with the Wald Z test. Mean comparisons were performed using the least significant difference for all significant effects (P ≤ 0.05).

The effect of the incubation animal (animal receiving incubation bags were deemed random blocks) and percent treatment difference upon the required sample size per treatment to obtain adequate statistical power for iDM and iNDF were simulated using a script developed for the R version 3.5.2 software (R Core Team, 2019). Data from both dietary treatments were pooled and analyzed for each method, iDM and iNDF, and sample type, feed and feces, combination. Percent treatment differences were determined from the pooled data set, and blocks (i.e., incubation animals) were assumed to span 2 to 16 in a research setting with the associated variance being determined for each combination from the pooled data. Fixed parameters for the simulation were as follows: α = 0.05 (statistical significance), β = 0.90 (statistical power), treatments = 8, blocks (i.e., incubation animals) ranging from 2 to 16, and treatment difference varying from 5% to 15% and 15% to 25% for feces and feeds, respectively.

Results

There was no interaction among BT × SA × IL (P > 0.05) for iDM and iNDF estimates or their resultant DMD and NDFD estimates. Data are presented for each significant 2-way interaction with main effects being discussed when a significant interaction was not present.

Basal Diet (Control)

Indigestible DM

Indigestible DM residue of the feed demonstrated an interaction of BT × SS with the F58 × 40 mg treatment having the greatest residue and the 40-mg rates demonstrating increased indigestible residue regardless of BT (Table 2). A BT × IL interaction for feed was present; the greatest residue remained in F58 × 288 h with both 576 h treatments having the least. There was no interaction of SS × IL for feed or feces with no demonstrated interaction of BT × SS or BT × IL for feces. Resultant DMD and NDFD demonstrated an interaction for BT × SS as the F58 × 20 mg treatment had greater digestibility estimates relative to all other combinations. The fecal residue was affected by the main effects of BT and SS with F58 and 40-mg treatments having greater iDM (Table 3). Incubation length had a marked effect upon DMD and NDFD as the 576-h treatment increased digestibilities 14% and 28%, respectively.

Table 2.

Two-way interactions for residues and digestibilities of iDM and iNDF for the high-concentrate diet without condensed tannins

		F57		F58		SEM, %	P-values3
Marker1	Item2	20	40	20	40		BT × SS
Bag type × sample size
Items
iDM	Feed residue, % DM	6.1c	6.6b	5.9c	7.2a	0.18	<0.01
	Fecal residue, % DM	17.8	19.4	19.3	20.2	1.04	0.36
	DMD, %	65.9b	66.9ab	70.1a	64.3b	1.91	<0.01
	NDFD, %	50.0b	51.1ab	55.8a	47.9b	3.42	0.01
iNDF	Feed residue, % DM	4.1b	4.4a	3.8c	4.5a	0.13	<0.01
	Fecal residue, % DM	13.0	13.9	13.7	14.5	0.72	0.66
	DMD, %	67.8b	68.5b	71.4a	67.7b	2.17	0.03
	NDFD, %	54.0b	55.0b	59.2a	54.1b	2.74	0.03
Bag type × incubation length
		F57		F58
		288 h	576 h	288 h	576 h		BT × IL
iDM	Feed residue, % DM	7.0b	5.8c	7.7a	5.4c	0.18	<0.01
	Fecal residue, % DM	18.6	18.5	19.9	19.6	1.04	0.73
	DMD, %	63.1	69.7	61.8	72.6	1.91	0.08
	NDFD, %	45.8	55.4	44.0	59.7	3.42	0.08
iNDF	Feed residue, % DM	4.9a	3.7b	5.0a	3.4c	0.13	<0.01
	Fecal residue, % DM	13.9	13.0	14.7	13.5	0.72	0.49
	DMD, %	64.8	71.5	65.3	73.7	2.17	0.39
	NDFD, %	49.7	59.3	50.6	62.7	2.74	0.39
Sample size × incubation length
		20		40
		288 h	576 h	288 h	576 h		SS × IL
iDM	Feed residue, % DM	6.9	5.1	7.8	6.1	0.18	0.58
	Fecal residue, % DM	18.4	18.7	20.1	19.5	1.04	0.24
	DMD, %	62.9	73.1	62.0	69.3	1.91	0.23
	NDFD, %	45.3	60.4	44.4	54.6	3.42	0.17
iNDF	Feed residue, % DM	4.7	3.3	5.1	3.8	0.13	0.71
	Fecal residue, % DM	13.8	12.9	14.9	13.5	0.72	0.24
	DMD, %	65.3	73.9	64.8	71.4	2.17	0.31
	NDFD, %	50.4	62.9	49.9	59.2	2.74	0.26

¹iDM, indigestible DM; iNDF, indigestible NDF.

²DMD, DM digestibility, NDFD, NDF digestibility.

³BT, bag type: 57 = 25-μm porosity and F58 = 10-μm porosity (ANKOM technology, Macedon, NY); SS, sample size: 20 and 40 mg/cm²; IL, incubation length: 288 and 576 h.

^a–dMeans followed by different superscripts within rows differ according to least significant difference (P ≤ 0.05).

Table 3.

Main effects for residues and digestibilities of iDM and iNDF for the high-concentrate diet without condensed tannins

		Main effects1
		Bag type		Sample size		Incubation length		SEM, %	P-value2
Marker3	Item4	F57	F58	20	40	288 h	576 h		BT	SS	IL
iDM	Feed residue, % DM	6.4	6.6	6.0	6.9	7.3	5.6	0.16	0.11	<0.01	<0.01
	Fecal residue, % DM	18.6	19.8	18.6	19.8	19.3	19.1	1.24	<0.01	<0.01	0.62
	DMD, %	66.4	67.2	68.0	65.6	62.4	71.2	1.71	0.52	0.06	<0.01
	NDFD, %	50.6	51.8	52.9	49.5	44.9	57.5	3.19	0.47	0.05	<0.01
iNDF	Feed residue, % DM	4.3	4.2	4.0	4.5	4.9	3.5	0.12	0.21	<0.01	<0.01
	Fecal residue, % DM	13.5	14.1	13.4	14.2	14.3	13.2	0.71	<0.01	<0.01	<0.01
	DMD, %	68.2	69.5	69.6	68.1	65.1	72.6	2.05	0.18	0.13	<0.01
	NDFD, %	54.5	56.7	56.6	54.5	50.2	61.0	2.55	0.14	0.15	<0.01

¹Bag type: 57, 25-μm bag and F58, 10-μm bag (ANKOM technology, Macedon, NY); sample size: 20 and 40 mg/cm²; incubation length: 288 and 576 h.

²BT, bag type; SS, sample size; IL, incubation length.

³iDM, indigestible DM; iNDF, indigestible NDF.

⁴DMD, DM digestibility, NDFD, NDF digestibility.

Indigestible NDF

An interaction of BT × SS was present for feed, as greater iNDF values were observed in the 40-mg treatment irrespective of BT with the F57 × 20 mg having the lowest residue (Table 2). The BT × IL interaction also influenced feed iNDF with the F58 × 576 h having the lowest iNDF residue. However, 288-h incubation had greater residue regardless of BT. As with iDM, there was no effect of SS × IL on iNDF of feed or feces with no interaction of BT × SS or BT × IL for feces. Digestibility estimates demonstrated an interaction of BT × SS for DMD and NDFD (Table 2). Use of F58 × 20 mg resulted in an average increase in digestibility of 4.9% and 8.9% for DMD and NDFD, respectively. The main effect of BT did affect the feces as F58 had elevated iNDF (Table 3). In contrast to iDM, SS and IL affected fecal residue as the 40-mg SS increased residue 6.1% and the 576-h IL reduced iNDF values 8.2%. Incubation length of 576 h resulted in an 11.6% increase in DMD and a 21.6% greater NDFD.

Condensed Tannin Diet

Indigestible DM

Feed and feces iDM were affected by BT × IL as the F58 × 288 h had the most residue with 576 h having the least irrespective of BT (Table 4). An interaction of SS × IL was present for the feed with 20 mg × 576 h having substantially less residue than the 40 mg × 288 h treatment. Sample size affected fecal residue as the 40-mg treatment corresponded to 8.6% greater residue (Table 5). Subsequent digestibilities did not display any interactions, but the 20-mg SS resulted in greater NDFD. As well, IL influenced both DMD and NDFD with the 576-h treatment increasing digestibility estimates 12% and 19%, respectively.

Table 4.

Two-way interactions for residues and digestibilities of iDM and iNDF for the high-concentrate diet with condensed tannins

		F57		F58			P-values1
Marker2	Item3	20	40	20	40	SEM, %	BT × SS
Bag type × sample size
Items
iDM	Feed residue, % DM	6.9	8.4	7.5	8.5	0.24	0.13
	Fecal residue, % DM	19.7	22.1	21.6	22.8	1.02	0.13
	DMD, %	65.0	62.5	66.6	62.0	2.92	0.58
	NDFD, %	54.6	51.1	56.5	50.3	5.3	0.55
iNDF	Feed residue, % DM	4.7	5.6	4.3	5.0	0.19	0.4
	Fecal residue, % DM	14.0	15.9	14.9	15.9	0.63	0.07
	DMD, %	64.0	62.0	68.7	64.8	1.86	0.52
	NDFD, %	50.6	50.2	58.6	53.3	4.6	0.21
Bag type × incubation length
		F57		F58
		288 h	576 h	288 h	576 h		BT × IL
iDM	Feed residue, % DM	8.2b	7.0c	9.2a	6.8c	0.24	<0.01
	Fecal residue, % DM	20.8b	21.0b	23.1a	21.3b	1.02	0.01
	DMD, %	60.4	67.1	60.4	68.3	2.92	0.74
	NDFD, %	48.6	57.1	48.5	58.2	5.30	0.80
iNDF	Feed residue, % DM	5.7a	4.6c	5.4b	3.9d	0.19	0.01
	Fecal residue, % DM	15.8	14.1	16.4	14.4	0.63	0.55
	DMD, %	62.0	63.9	63.1	70.4	1.86	0.08
	NDFD, %	48.2	52.6	51.4	60.5	4.60	0.21
Sample size × incubation length
		20		40
		288 h	576 h	288 h	576 h		SS × IL
iDM	Feed residue, % DM	7.8b	6.5d	9.6a	7.3c	0.24	<0.01
	Fecal residue, % DM	20.7	20.6	23.2	21.7	1.02	0.07
	DMD, %	63.1	68.6	57.7	66.8	2.92	0.34
	NDFD, %	52.4	58.7	44.7	56.7	5.30	0.22
iNDF	Feed residue, % DM	5.1b	3.9d	6.1a	4.5c	0.19	0.03
	Fecal residue, % DM	15.3	13.7	16.9	14.8	0.63	0.19
	DMD, %	64.5	68.2	60.6	66.1	1.86	0.54
	NDFD, %	51.6	57.6	48.0	55.5	4.60	0.70

¹BT, bag type: 57, 25-μm porosity and F58, 10-μm porosity (ANKOM technology, Macedon, NY); SS, sample size: 20 and 40 mg/cm²; IL, incubation length: 288 and 576.

²iDM, indigestible DM; iNDF, indigestible NDF.

³DMD, DM digestibility; NDFD, NDF digestibility.

^a–dMeans followed by different superscripts within rows differ according to least significant difference (P ≤ 0.05).

Table 5.

Main effects for residues and digestibilities of iDM and iNDF for the high-concentrate diet with condensed tannins

		Main effects1
		Bag type		Sample size		Incubation length		SEM (%)	P-value2
Marker3	Item4	F57	F58	20	40	288 h	576 h		BT	SS	IL
iDM	Feed residue, % DM	7.6	8.0	7.2	8.4	8.7	6.9	0.21	0.03	<0.01	<0.01
	Fecal residue, % DM	20.9	22.2	20.6	22.4	21.9	21.2	0.97	<0.01	<0.01	0.07
	DMD, %	63.7	64.3	65.8	62.2	60.4	67.7	2.61	0.76	0.06	<0.01
	NDFD, %	52.8	53.4	55.5	50.7	48.5	57.7	5.04	0.81	0.04	<0.01
iNDF	Feed residue, % DM	5.1	4.7	4.5	5.3	5.6	4.2	0.18	<0.01	<0.01	<0.01
	Fecal Residue, % DM	14.9	15.4	14.5	15.9	16.1	14.2	0.61	0.06	<0.01	<0.01
	DMD, %	63.0	66.7	66.3	63.4	62.6	67.1	1.45	0.01	0.06	<0.01
	NDFD, %	50.4	55.9	54.6	51.7	49.8	56.5	4.39	<0.01	0.14	<0.01

¹Bag type: 57, 25-μm bag and F58, 10-μm bag (ANKOM technology, Macedon, NY); sample size: 20 and 40 mg/cm²; incubation length: 288 and 576 h.

²BT, bag type; SS, sample size; IL, incubation length.

³iDM, indigestible DM; iNDF, indigestible NDF.

⁴DMD, DM digestibility; NDFD, NDF digestibility.

Indigestible NDF

Similar to the basal diet, iNDF of the feed demonstrated a BT × IL interaction with F58 × 576 h having the least iNDF residue; however, in contrast to the basal diet, feed iNDF was only elevated in the F57 × 288-h treatment (Table 4). An interaction of SS × IL was present for the feed, trending similarly to iDM as the 40 mg × 288 h had the greatest residue and 20 mg × 576 h the least. Indigestible NDF of the feces was affected by the main effects, SS and IL, as the 20-mg and 576-h treatments had 8.6% and 11.6% less iNDF residue, respectively (Table 5). Similar to iDM, digestibility estimates were only influenced by the main effects. Greater DMD and NDFD were observed when using the F58 bag, +5.9% and 10.9%, respectively, and when the 576-h IL was utilized, +7.3% and 13.5%, respectively. A summary of P-values for main effects and all interactions for both diets are presented in Table 6.

Table 6.

P-values for main effects and all interactions

		P-value1
Marker2	Item3	BT	SS	IL	BT × SS	BT × IL	SS × IL	BT × SS × IL
Basal diet
iDM	Feed residue	0.11	<0.01	<0.01	<0.01	<0.01	0.58	0.38
	Fecal residue	<0.01	<0.01	0.62	0.36	0.73	0.24	0.60
	DMD	0.52	0.06	<0.01	<0.01	0.08	0.23	0.62
	NDFD	0.47	0.05	<0.01	0.01	0.08	0.17	0.70
iNDF	Feed residue	0.21	<0.01	<0.01	<0.01	<0.01	0.71	0.51
	Fecal Residue	<0.01	<0.01	<0.01	0.66	0.49	0.24	0.90
	DMD	0.18	0.13	<0.01	0.03	0.39	0.31	0.46
	NDFD	0.14	0.15	<0.01	0.03	0.39	0.26	0.51
Basal diet + CT
iDM	Feed residue	0.03	<0.01	<0.01	0.13	<0.01	<0.01	0.27
	Fecal residue	<0.01	<0.01	0.07	0.13	0.01	0.07	0.28
	DMD	0.76	0.06	<0.01	0.58	0.74	0.34	0.57
	NDFD	0.81	0.04	<0.01	0.55	0.80	0.22	0.33
iNDF	Feed residue	<0.01	<0.01	<0.01	0.40	0.01	0.03	0.59
	Fecal Residue	0.06	<0.01	<0.01	0.07	0.55	0.19	0.34
	DMD	0.01	0.06	<0.01	0.52	0.08	0.54	0.58
	NDFD	<0.01	0.14	<0.01	0.21	0.21	0.70	0.60

¹BT, bag type; SS, sample size; IL, incubation length.

²iDM, indigestible DM; iNDF, indigestible NDF.

³DMD, DM digestibility; NDFD, NDF digestibility.

When comparing CV, within-sample types, it is evident that iNDF has less variation relative to iDM (Fig. 1). Overall variability was greater for the diet containing CT than for the basal diet, regardless of the marker. Feces displayed much less volatility as both diets had a very similar CV with a consistent reduction in variation from iDM to iNDF. Independent of diet, the average CV of feed was 29.3% and 26% for iDM and iNDF, respectively, whereas the corresponding CV of the feces were 26.3% and 21.5%.

Figure 1.

Coefficients of variation for indigestible residues within sample types and diets. (A, Feed, diet without condensed tannins; B, Feed, diet with condensed tannins; C, Feces, diet without condensed tannins; D, Feces, diet with condensed tannins; , indigestible DM; , indigestible NDF). BT, Bag type, SS, Sample size, IL, Incubation length.

Variances and Sample Size

Variance partitioning of iDM residue indicated that incubation animal (cannulated animal) had only a small effect on residue variability, although the effect was greater upon feces than feed samples (Table 7). The inverse occurred for iNDF as 3% to 4% of the variation of feed was dependent on the incubation animal. When exploring the partitioning for digestibilities, trial animal accounted for a substantial portion of the variability of DMD, whereas collection period (period sample was collected in the previous trial) accounted, on average, for 46.5% of the variation associated with NDFD when iDM was used as an internal marker. The variance partitioning for iNDF residues was inverse of the trend illustrated by iDM as the variability of incubation animal increased within the feed component and decreased in the feces. Partitioning within digestibilities proved volatile as trial animal accounted for a minor portion of the variation in comparison to collection period for DMD of the basal diet and NDFD of the CT diet. By contrast, trial animal represented the major source of variability relative to the collection period for both DMD of the CT diet and NDFD of the basal diet. Reviewing the sample size simulations (Fig. 2), within our trial, feed required far fewer treatment replications to obtain 90% power than feces.

Table 7.

Variance partitioning of random effects

Marker1	Item2	Incubation animal	Animal	Collection period3	Residual
Basal diet
iDM	Feed residue	1.1%	—	—	98.9%
	Fecal residue	2.8%	—	—	97.2%
	DMD	—	43.8%	0.0%	56.2%
	NDFD	—	5.7%	43.8%	50.5%
iNDF	Feed residue	4.1%	—	—	95.9%
	Fecal residue	0.6%	—	—	99.4%
	DMD	—	6.5%	50.1%	43.3%
	NDFD	—	45.9%	2.1%	52.0%
Basal diet + condensed tannin
iDM	Feed residue	0.7%	—	—	99.3%
	Fecal residue	2.1%	—	—	97.9%
	DMD	—	19.8%	15.6%	64.6%
	NDFD	—	9.9%	49.2%	40.9%
iNDF	Feed residue	3.3%	—	—	96.7%
	Fecal residue	0.0%	—	—	100.0%
	DMD	—	7.4%	0.8%	91.8%
	NDFD	—	15.1%	47.4%	37.5%

¹iDM, indigestible DM; iNDF, indigestible NDF.

²DMD, DM digestibility; NDFD, NDF digestibility.

³Collection period, period of sample collection within digestibility trial.

Figure 2.

Estimated sample size per treatment, number of blocks, and percent treatment difference required to detect a statistical difference among treatments for indigestible components (A, indigestible dry matter feces, B, indigestible dry matter feed, C, indigestible neutral detergent fiber feces, D, indigestible neutral detergent fiber feed), assuming a power of 90% and an α-value of 5%. The area below the shaded plane indicates differences capable of being detected with the sample size used in the present study. The sample size and treatment differences for indigestible feed residues exceeded requirements for the detection of differences within this simulation.

Discussion

The utilization of iDM, rather than indigestible residue components, as an internal marker would be advantageous due to a reduction in analytical procedures (Detmann et al., 2001). However, microbial and non-microbial contamination from ruminal incubation can introduce large, heterogeneous variability (Huhtanen et al., 1994; Valente et al., 2011), reducing the precision of estimates except for situations in which iNDF is a large component of the iDM. Within the present study, a reduction in contaminants following washing in ND solution was evident when comparing iDM and iNDF values of the respective sample and diet combinations. Correspondingly, all iNDF residues demonstrated lower CV than those of iDM with an average reduction of 3.5% and 4.8% for feed and feces, respectively. Greater variation is commonly observed for iDM relative to iNDF (Valente et al., 2011); although the accuracy of recovery estimates is not commonly affected, the precision is typically reduced when using iDM (Sampaio et al., 2011). Within our study, there was large variation for feed containing CT due to study treatments, whereas feces from the CT diet was less variable. The CT present within feed commonly has greater precipitation capacity compared with those in feces (unpublished data), thereby enabling greater inhibition of microbes and associated enzymes, as well as the formation of CT complexes with endogenous/exogenous substrate and microbes (Haslam, 1989). Reduction in precision due to the presence of secondary compounds could potentially influence markers for estimation of intakes, excretion, and digestibilities. Tedeschi and Fox (2018) noted that discrepancies in iNDF greatly influenced TDN values, with tropical feedstuffs commonly being penalized to a greater extent than temperate. This could be a consequence of antinutritional factors, such as CT, being more widely distributed in tropical species (Muir et al., 2009).

In general, DMD estimations from iNDF exceeded those of iDM by 0.8% and 1.9%, on average, for the basal and CT diet, respectively. This is consistent with Huhtanen et al. (1994) who showed that iDM provided lower estimates of DMD when compared with iNDF. The NDFD of the basal diet had an average change of +8.1% when using iNDF, whereas NDFD of the CT diet only exhibited a +0.07% change on average when calculated from iNDF. Comparing iDM and iNDF, digestibilities of the CT diet were variable with differences in DMD and NDFD ranging from −3.4% to 3.4% and −4.7% to 3.7%, respectively. It is likely that greater variability and reduced differences in digestibilities are not solely due to indigestible nutritional constituents, but rather an artifact of insoluble CT–substrate complexes due to tannins being readily confounded with indigestible components (Van Soest, 2015). A major issue with the presence of CT is that the ND washing process provides an environment, high heat and pH 7, that induces additional polymerization and does not break CT–substrate complexes, rather there is an increase of insoluble CT complexes (Van Soest, 1994). Due to the variability present within residues and digestibilities of CT samples, when analyzing a sample containing CT, use of multiple sequential washes may be a method of accounting for CT–complex retention for improved precision (Van Soest and Robertson, 1985; Van Soest, 1994).

Collectively, the F58 bags resulted in greater iDM residue values for all samples. However, because a distinction between iDM and exogenous material cannot be made the effect upon an incubated sample is unknown, but increased contamination was observed with the F58 bags due to material characteristics. A greater percent change following washing with ND solution was observed in F58 bags within BT × SS and BT × IL treatments. This is further evidenced by the change in correction factor of blank bags following washing with ND solution. The F58 correction factor decreased 4.5% and 5.8% for 288- and 576-h IL, respectively, whereas the corresponding F57 correction factor only decreased 1.4% and 1.9%. The use of lower porosity bags (<10 μm) for in situ methods can result in estimation errors due to reduced microbial influx and micro-environments within the bag (Nocek, 1988; Huhtanen et al., 1998); however, fermentation did not appear to be hindered due to bag porosity in our study as feedstuffs within F58 bags had equal or lesser iNDF residues. Because bag porosity did not affect degradation, it appears probable that there was sample washout for feces incubated within F57 bags as evidenced by lower fecal recoveries relative to the F58. This finding is consistent with the greater fecal loss with increased bag porosity noted by Huhtanen et al. (1994).

Sample size had a substantial effect upon fecal residues regardless of marker or diet, with 40 mg resulting in greater indigestible constituents relative to contemporaries. Both BT × SS and SS × IL interactions for feeds demonstrated a greater percent change from iDM to iNDF when utilizing the larger SS, indicative of soluble material being removed during washing with ND. Larger SS reduced pore size-to-free surface area ratio that likely resulted in reduced microbial and enzymatic activity, decreasing the rate of degradation (Huhtanen et al., 1998; Vanzant et al., 1998). This would more greatly affect bags containing larger amounts of fermentable substrate that are exposed to the rumen environment for a shorter period of time or are predisposed to higher contamination levels. Although the use of 10 mg/cm² is commonly utilized for in situ studies (Vanzant et al., 1998), and may provide more accurate estimates of truly indigestible components, small residues following extended incubation length could potentially introduce more error due to greater analytical precision required (Sampaio et al., 2011).

The 576-h IL tended to result in less residue, even in the presence of contaminants. For SS × IL interaction, the larger SS resulted in elevated residues for both IL, whereas within the BT × IL interaction, less feed residue remained for 576-h incubations relative to the 288 h. Within a diet, iDM was similar irrespective of BT when incubated for 576 h; however, F58 × 576 h had lower iNDF residue than the F57. The difference in BT for iNDF is likely a result of greater bag contamination that was removed during the ND wash. In general, the 576-h incubations exhibited lower, more homogenous estimates of residue that yielded substantially higher digestibility approximations. This is in agreement with Koukolová et al. (2004) who reported considerably greater degradation when using 504-h incubations vis-á-vis 168 h. The 576-h incubations appear to better represent the indigestible constituents as digestibility estimates using iDM and iNDF varied only slightly in most instances, whereas 288-h estimates demonstrated variability indicative of digestible material removal during ND wash. However, an increase in contamination with prolonged IL was evident as the percent change from iDM to iNDF was generally lower for 288- versus 576-h incubations. Due to increased contamination, caution must be taken when utilizing iDM. In this particular study, 288-h incubations do not appear adequate for the determination of truly indigestible constituents or precise digestibility estimations when utilizing 20 and 40 mg/cm² SS-to-surface area ratio.

During this trial, the CV across treatment combinations was elevated, exceeding 30% in some scenarios. However, previous research investigating iNDF of various feeds has noted high CV (>25%) within individual feedstuffs (Casali et al., 2009; Valente et al., 2011; Raffrenato et al., 2018). When comparing BTs and particle sizes, Valente et al. (2011) observed CV ≥ 25% for corn gluten meal, corn grain, corn silage, straw, sorghum grain, and soybean meal. Across trials, soybean hulls displayed CV ranging from 26% to 34% (Casali et al., 2009; Valente et al., 2011). Evaluation of forages at different physiological stages resulted in CV ≥ 25% for alfalfa, barley, corn, oats, ryegrass, and sorghum (Raffrenato et al., 2018). The Nordic ringtest for iNDF observed variation across laboratories as the CV for barley, grass silage, and rapeseed meal were 18%, 23%, and 25%, respectively (Lund et al., 2004). This was largely an effect of variation in BT, length of incubation, incubation animal, and washing procedures of the laboratories. Although the Nordic ringtest compared laboratories, it provides insight into how minor alterations in methodology can result in large discrepancies. Relative to the variation observed in previous research, the CV of feed and feces in the current trial are in no manner extreme based on the factorial treatment arrangement implemented.[AU: The following are the standard abbreviations that can be used without definition except at the beginning of the sentence: SEM, DM, CP, ADF, NDF, CV, ME. Your paper has been amended accordingly. Please check.]

For the current simulation, we used an average sample size of 90 bags per treatment with an average treatment difference of 22.5% for iDM and iNDF feed residues, exceeding the requirements to detect a true difference if present. However, for feces, our design did not enable 90% statistical power to be attained due to treatment differences ranging from roughly 8% to 10%. Based on our data, if a 25% difference between incubation treatments is sought or anticipated and only 2 incubation animals are available, a total of 24 bags per treatment (12 bags per incubation animal) is required. This level of replication greatly exceeds the triplicate and quadruplicate replications per incubation animal that is often used for in situ experimentation. The required sample size for adequate statistical power was greatly affected by sample type, as a reduction in treatment differences within the feces restricted the power obtained relative to that of the feed. When comparing iDM and iNDF of feces, iNDF reduced the sample size required by decreasing incubation animal variation and increasing treatment differences. For feed residue, increased sample size is needed for iNDF than that of iDM, due to greater incubation animal variance and lesser treatment differences. The divergence between feed and feces is a consequence of the total fermentable substrate present within each sample type. This is evidenced by less digestible feeds commonly having reduced variation relative to more digestible feedstuffs (Valente et al., 2011; Raffrenato et al., 2018). The use of a concentrate diet within our study provided a large amount of potentially fermentable substrate that enabled larger treatment differences to be observed; however, when investigating methodologies or comparing feedstuffs, greater sample size (replication) should be utilized for determination of indigestible contents of roughage diets or feces due to the lower total fermentable substrate that could result in less prominent treatment differences. Although the percent difference among treatments had the greatest effect upon attaining adequate sample size, the number of blocks (i.e., incubation animal) introduced variation that could have a substantial impact on results and applicability. Incubation animal variation was greatest for feed iNDF, indicating that individual incubation animals can influence the overall degradation, resulting in less control of the variance. Therefore, the use of iNDF strictly as a marker when feeding similar diets, a lesser number of incubation animals (2 to 4) may reduce the overall variation; however, if iNDF is to be applied within nutritional models, greater incubation animal numbers are required (≥4) so that biological variation may be encompassed as best possible.

In summary, our results indicate less error and greater precision in the determination of indigestible components using iNDF. Estimates of iDM are largely influenced by contamination that can vary greatly based on intrinsic factors associated with the bag and incubation duration. However, the presence of CT appears to influence both iDM and iNDF to varying degrees. In agreement with Krizsan et al. (2015), in situ determination of indigestible residues should be performed using bags with a pore size ranging from 6 to 12 μm as the use of F57 bags (25 μm) resulted in lower fecal sample recovery when compared with F58 bags (10 μm). Within this study, SS-to-surface area ratio of 20 mg/cm² is best suited for determination of indigestible components as 40 mg/cm² appears inhibitory to degradation, promoting larger residues that were not corrected when using 576-h incubations. However, the determination of indigestible residues utilizing 10 mg/cm² could potentially reduce the incubation length required and the accompanying increase in contamination, but error associated with analytics must be taken into account. Lower indigestible residues resulted from 576-h incubations relative to those of 288 h. This indicates that pdNDF is still available after 288 h within the rumen, albeit a function of IL and SS. In this study, improved precision in the estimation of indigestible components was attained through the determination of iNDF using F58 bags, 20 mg/cm² SS, and 576-h IL. As total fecal collection was not performed, we could not determine percent recovery or compare digestibility estimates with in vivo data. Therefore, although this treatment combination improved estimate precision, the accuracy of the estimate requires further investigation. In actuality, the use of different BTs and incubation lengths for individual sample forms (feed and feces) may provide the greatest combination of accuracy and precision, but this needs to be examined more in-depth. When utilizing indigestible components, the determination of associated power for a specific sample type, and the goal of the specific study is a prerequisite. The assumption that 2 incubation animals and low treatment replication will suffice appears inappropriate for indigestible residues, as sample numbers need to be increased 3- to 4-fold when utilizing similar sample types to obtain adequate power for determination of methodological or dietary differences. Further research should investigate the impact of secondary compounds beyond CT upon accuracy and precision of estimates attained using internal markers. A better understanding of SS × IL combination dynamics is needed prior to the development of a standardized methodology.