Beruflich Dokumente
Kultur Dokumente
PII: S0737-0806(17)30497-5
DOI: 10.1016/j.jevs.2017.08.015
Reference: YJEVS 2373
Please cite this article as: Hansen TL, Lawrence LM, Composition factors predicting forage digestibility
by horses, Journal of Equine Veterinary Science (2017), doi: 10.1016/j.jevs.2017.08.015.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
1
PT
5
RI
6
SC
8
10
U
AN
11
12 Department of Animal and Food Sciences, University of Kentucky, Lexington, KY 40546 USA
M
13
D
14
TE
15
16
EP
*
17 Corresponding author. Tel: 859-257-7509; Fax: 859-257-2534; EM: llawrenc@uky.edu
18
C
AC
ACCEPTED MANUSCRIPT
2
19 Abstract
20 Forages play a key role in the equine diet; however, assessing the feeding value of
21 forages through in vivo digestibility studies is costly and time-consuming. The ability to evaluate
22 feeding value through a model relating chemical composition to forage digestibility could be a
PT
23 useful tool in the equine industry. The objective of this study was to relate forage chemical
RI
24 composition to in vivo dry matter digestibility (DMD) using previously published data, with the
25 expectation that forage DMD would be most significantly related to neutral detergent fiber
SC
26 (NDF) and acid detergent fiber (ADF) content of the forage. Simple linear regression results
27 indicated more variation in forage digestibility was explained by NDF and crude protein (CP)
28
U
concentrations than ADF (r2 = 0.6017, 0.5402, and 0.4643 respectively). The best 2 variable
AN
29 model for equine forage DMD was related to NDF and CP concentrations (DMD = 65.81 +
30 0.7207×CP – 0.3514×NDF; DM basis; P<0.0001, R2 = 0.6690, adj. R2 = 0.6583). Model fit was
M
31 marginally improved with the inclusion of 3 predictor variables. Comparison of these equations
D
32 to independent data sets not used in model development showed promise in predicting forage
TE
33 digestibility within the range of forages used to generate the equations (CP: 2.9 – 22.6, ADF:
34 25.4 – 52.4, NDF: 32.8 – 83.3, DM basis). Further study is needed to expand the equations for a
EP
35 broader range of forages, but equine professionals and horse owners can use NDF and CP to
36 make relative comparisons of forage value with typical hays fed to horses.
C
38 Abbreviations: acid detergent fiber (ADF), adjusted multiple coefficient of determination (adj.
39 R2), coefficient of determination (r2), crude fiber (CF), crude protein (CP), digestible dry matter
41 determination (R2), neutral detergent fiber (NDF), organic matter (OM), percent (PER),
ACCEPTED MANUSCRIPT
3
42 predictive coefficient of variation (pred. R2), predictive residual sum of squares (PRESS),
PT
RI
U SC
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
4
44 1. Introduction
45 Under most conditions, forage constitutes an important part of the equine diet.
46 Digestibility is one of the factors that affects the feeding value of forage. The gold standard for
PT
48 vivo digestibility is costly, labor intensive, and returns information slowly. Thus, there has been
RI
49 considerable interest into alternative methods to evaluate forage quality such as in situ and in
50 vitro assays [1, 2], but these assays are also relatively slow to return information.
SC
51 Ruminant nutritionists and forage scientists have used forage chemical composition to
52 develop indices such as relative feed value (RFV) and relative forage quality (RFQ) [3, 4].
53
U
These indices are largely based on NDF and ADF concentrations, with greater NDF and ADF
AN
54 concentrations indicating lower forage quality. These indices provide a virtually instantaneous
55 evaluation of forage quality for ruminants; however, it is unclear how these indices apply to
M
56 horses. While there have been many studies on factors affecting digestibility of feeds by horses,
D
57 studies that have attempted to assess the relationships between forage chemical composition and
TE
58 forage digestibility by horses have been limited by number or type of forages utilized [5].
59 Broader datasets are needed to clarify the relationships between chemical composition and
EP
61 The objectives of this study were to determine if chemical composition could be used to
C
62 predict forage digestibility by horses and if so, to develop a model for predicting DMD. Rather
AC
63 than conduct additional in vivo experiments, we utilized data from previously published studies
64 representing a wide range of forages. We hypothesized that forage NDF and ADF
65 concentrations would be the most useful predictors of forage digestibility, and with increasing
69 In order to obtain a large range of values for in vivo forage digestibility by horses, a
70 literature search was conducted utilizing the search engine Google Scholar using the keywords
PT
71 “digestibility”, “horses”, and “forage”. Articles and abstracts were selected based on the use of
RI
72 total fecal collections in mature horses or ponies fed all forage diets. Studies using internal or
73 external markers, young horses (< 2 y of age), or those determined to have non-standard
SC
74 practices (use of cannulated horses, short dietary adaptation, or inadequate description of
75 calculations or methods) were excluded from analysis. Studies were included if a portion of the
76
U
study fit all of the criteria (Example: For a study comparing forage digestibility between growing
AN
77 and mature horses, only the mature horse data were retained for analysis). Studies must have
78 reported relevant chemical components (NDF, ADF, and CP) as well as DMD. Additionally, the
M
79 same criteria were applied to the proceedings from the Equine Nutrition and Physiology
D
80 Symposium and Equine Science Symposium from 1968 through the year 2011. The literature
TE
81 search resulted in 26 studies that fit the specified criteria (Table 1).
82 The data collected from these studies are shown in Table 2. The amount of hemicellulose
EP
83 (HEM) in the forage was estimated by HEM = NDF-ADF. The percent of ADF in the NDF
84 portion of the forage (PER) was calculated as PER = ADF/NDF×100. In total, there were 70
C
85 DMD mean values from 70 forage diets. There were 65 measurements of NDF concentrations,
AC
87 measurements of HEM, and 65 measurements of PER. There were 65 DMD measurements with
89 concentrations).
ACCEPTED MANUSCRIPT
6
90 The mathematical relationships between diet digestibility and forage composition using
91 data from the selected studies were evaluated utilizing both simple and multiple linear
92 regression. Explanatory variables included NDF, ADF, and CP concentrations of the forages in
93 addition to calculated HEM and PER values. Multiple linear regression used combinations of
PT
94 these explanatory variables. Polynomial regression included the square of explanatory variables
RI
95 in multiple regression models. A classification variable for type of forage was not included due
96 to the small number of available forage diets for certain types of forages (only one measurement
SC
97 for warm-season legume forages, and less than 15 complete measurements each for cereal grain
98 hays, cool-season grasses, and cool-season legumes). In addition, a classification variable for
99
U
methods used to analyze forage chemical composition was not included due to insufficient
AN
100 information in some papers.
102 Regression analysis was performed with the PROC REG procedure of SAS software
D
103 (Version 9.3; SAS Institute, Inc., Cary, NC). Simple linear regression models were compared
TE
104 using the coefficient of determination (r2), whereas multiple linear regression models were
105 evaluated based on the adjusted R2 (adj. R2). The use of the adj. R2 in multiple regression
EP
106 models was necessary in order to account for a continual increase in the multiple coefficient of
107 determination when additional explanatory variables were added to the model.
C
108 Regression models were tested with data from three equine digestibility studies (4
AC
109 different forages; Table 3) not included in the initial regression analysis data [31-33]. A paired t-
110 test was used to compare calculated DMD to in vivo values. Predictive ability of models was
111 determined by predicted residual sum of squares (PRESS) and predictive R2 (pred. R2) statistics.
112 3. Results
ACCEPTED MANUSCRIPT
7
113 Simple linear regression indicated all forage composition factors were significant to
114 predict in vivo DMD (Table 4); however, the best fit of the available data was achieved when
115 using NDF concentrations of forages (r2 = 0.6017; Figure 1). Crude protein concentrations
116 showed the second best fit to the available data (r2 = 0.5402). Less than 50% of the variation in
PT
117 DMD was explained by the ADF concentration of the forage (r2 = 0.4643; Figure 2).
RI
118 When using two predictor variables for multiple linear regression, the best fit for forage
119 composition to predict in vivo DMD was achieved (as determined by the adj. R2) with the
SC
120 inclusion of the predictor variables NDF and CP (Table 5). Dry matter digestibility was related
121 to NDF and CP concentrations by DMD = 65.81 + 0.7202×CP – 0.3514×NDF (DM basis;
122
U
P<0.0001, R2 = 0.6690, adj. R2 = 0.6583). The addition of any third forage composition
AN
123 explanatory variable to that model did not increase the fit of the model as determined by the adj.
124 R2 nor were all three explanatory variables individually significant. However, formation of other
M
125 three variable models increased the fit of the model to the data available. The model of best fit
D
126 was achieved with CP, ADF, and PER predictor variables (DMD = 56.16 + 0.6080×CP –
TE
128 Polynomial variables were not significant in the models when using one forage
EP
129 composition variable (for example, relating DMD to NDF and NDF2). Inclusion of multiple
130 composition variables in polynomial regression did not increase the fit of the model, nor were all
C
132 Using the independent data from Table 3, predictive ability was determined for the three
133 best fitting regression models (bolded in Table 5). Predictive statistics were marginally improved
134 for the models relating DMD to CP, ADF, and HEM (PRESS = 2012.55, pred. R2 = 0.6426) and
135 CP, ADF, and PER (PRESS = 2008.01, pred. R2 = 0.6434) compared to the two predictor model
ACCEPTED MANUSCRIPT
8
136 (PRESS = 2053.35, pred. R2 = 0.6353). There was no difference between observed in vivo
137 digestibility and model predicted digestibility (P> 0.1; Figure 3).
138 4. Discussion
139 Although all composition factors had a significant influence on DMD, this study found
PT
140 the best predictive factor for DMD was NDF concentration when using simple linear regression.
RI
141 Martin-Rosset et al. [5] reported digestible organic matter (OM) was related to CP, crude fiber
142 (CF), NDF, and ADF concentrations of legume and grass hays, again demonstrating the best
SC
143 linear fit was achieved when using NDF concentration to predict forage digestibility by horses
144 (r2 = 0.707 in that study). The coefficients of determination for all comparable equations were
145
U
greater in that study than the current study. Martin-Rosset et al. [5] may have observed a better
AN
146 fit because those studies were all conducted at the same institution allowing for more consistency
147 of study methods. In the present study there were differences in analytical procedures as well as
M
148 fecal collection procedures among studies. Additionally, in that study [5] most of the forages
D
149 were cool-season grass hays, whereas this study uses a broader range of forage types.
TE
150 The relatively low coefficient of determination for ADF concentration to predict in vivo
151 DMD was unexpected because ADF has often been used to determine forage digestibility by
EP
152 ruminants. For example, the recommended equation to determine digestible DM (DDM) when
153 calculating RFV of a forage is DDM = 88.9 – 0.770×ADF (DM basis) [34]. However, it is
C
154 important to consider that DDM equation has been shown to poorly predict DDM in some
AC
155 instances [4]. Rohweder et al. [3] also found forage ADF concentrations were useful to predict
156 alfalfa DMD, but the fit was poorer with grasses than legumes. Forage ADF concentrations may
157 not have fit the data well in the current study because the dataset used a wide variety of forages,
158 including warm- and cool-season grasses and legumes. Also, ADF concentrations had a
ACCEPTED MANUSCRIPT
9
159 relatively small range of values compared to NDF and CP and may have led to the poor fit
161 Multiple linear regression with forage composition variables NDF and CP achieved a
162 better fit to the data than simple linear regression. The use of a third forage composition
PT
163 explanatory variable marginally increased the fit of the model, but in many instances all three
RI
164 explanatory variables in those models were not statistically significant. This observation could be
165 due to a correlation among composition variables within the forages. Models that performed well
SC
166 included CP and fiber composition variables. The models with the highest fit explained
167 approximately two-thirds of the variation associated with digestibility. It would be unlikely that
168
U
forage chemical composition alone would explain all of the variation in digestibility as other
AN
169 factors such as level of intake and degree of processing could also influence digestibility.
170 Using the 3 bolded multiple regression equations in Table 5, the differences between the
M
171 observed and predicted values for digestibility of forages in the independent data sets were
D
172 minimal. Therefore, these equations show promise for estimating forage feeding value for
TE
173 horses. Only slight improvements were observed with the use of a multiple regression equation
174 with 3 dependent variables, indicating the two variable model relating NDF and CP
EP
175 concentrations to DMD is sufficient for assessing feeding value. More independent data sets are
176 needed in order to test the models for a range of forage composition values; however, this step is
C
177 currently limited by the availability of additional applicable studies. Also, all of the forages used
AC
178 in the present study were dry, so additional studies are needed to examine the applicability of the
180 Agreement between predicted and observed values could potentially be improved if
181 equations could be made specific for the type of forage (cool-season grasses, warm-season
ACCEPTED MANUSCRIPT
10
182 grasses, legumes, etc.) as has been done in ruminant studies [3, 35]. However, categorical
183 analysis was not practical in this assessment due to the small sample size from the literature.
184 Once additional digestibility studies using pure legume or pure grass forages are published the
185 feasibility of equations for specific forage type (or genus) can be revisited. However, horses are
PT
186 often fed mixed hays containing both legumes and grasses; therefore, a model that predicts
RI
187 across multiple forage types could have practical value in the horse industry.
188 5. Conclusions
SC
189 With a large range of in vivo DMD values from published equine forage digestibility
190 research, regression results indicated digestibility was related to forage chemical composition.
191
U
Overall, forage digestibility was negatively correlated to fiber concentrations and positively
AN
192 correlated to CP concentrations. Crude protein and NDF concentrations can be used to quickly
193 assess forage composition and determine more digestible forages, and therefore, forages with
M
194 greater feeding value. Further work is needed to determine if this model is applicable to forages
D
195 with nutrient compositions beyond the range of forages used to create the models. Individual
TE
196 models for different forage types may better predict DMD, but the limited number of equine
197 digestibility studies does not allow for that analysis at this time. The current model has
EP
198 advantages of easily determined nutrient composition factors and a wide range of applicability.
199 Acknowledgements
C
200 This is publication 15-07-035 of the Kentucky Agricultural Experiment Station and is
AC
201 published with the approval of the Director. This work is supported by the National Institute of
203 Highlights
204 • Forage chemical composition can be used to estimate forage quality and digestibility.
ACCEPTED MANUSCRIPT
11
205 • More variation in forage digestibility was explained by NDF and CP values than ADF.
206 • Addition of a third regression variable marginally increased model predictive ability.
208
PT
RI
U SC
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
12
209 References
210 1. Koller BL, Hintz HF, Robertson JB, Van Soest PJ. Comparative cell wall and dry matter
211 digestion in the cecum of the pony and the rumen of the cows using in vitro and nylon
PT
213 2. Earing JE, Cassill BD, Hayes S.H, Vanzant E., Lawrence LM. Comparison of in vitro
RI
214 digestibility estimates using the DAISYII incubator with in vivo digestibility estimates in
SC
216 3. Rohweder DA, Barnes RF, Jorgensen N. Proposed hay grading standards based on laboratory
217 analyses for evaluating quality. J Anim Sci 1978; 47: 747-59.
218
U
4. Moore J, Undersander D. Relative Forage Quality: An alternative to relative feed value and
AN
219 quality index, in: Proc 13th Florida Rumininant Nutr Symp Gainesville FL 2002; pp 16-
220 31.
M
222 matter digestibility of forages in horses using different chemical, biological and physical
TE
223 methods, in Saastamoinen, M., Fradinho, M.J., Santos, A.S., Miraglia, N. (Eds.), Forages
226 6. Aiken GE, Potter GD, Conrad BE, Evans JW. 1989. Voluntary intake and digestion of Coastal
C
227 bermudagrass hay by yearling and mature horses. J Equine Vet Sci 1989; 9: 262-64.
AC
229 determined by total collection of faeces and acid-insoluble ash an internal marker. Livest.
231 8. Crozier JA, Allen VG, Jack NE, Fontenot JP, Cochran MA. Digestibility, apparent mineral
232 absorption, and voluntary intake by horses fed alfalfa, tall fescue, and Caucasian
234 9. Cuddeford D, Pearson RA, Archibald RF, Muirhead RH. Digestibility and gastrointestinal
PT
235 transit time of diets containing different proportions of alfalfa and oat straw given to
RI
236 Thoroughbreds, Shetland ponies, Highland ponies and donkeys. Brit Soc Anim Nutr
SC
238 10. Cymbaluk NF, Christensen DA. Nutrient utilization of pelleted and unpelleted forages by
240
U
11. Cymbaluk NF. Comparison of forage digestion by cattle and horses. Can J Anim Sci 1990;
AN
241 70: 601-10.
242 12. Drogoul C, Poncet C, Tisserand J. Feeding ground and pelleted hay rather than chopped hay
M
243 to ponies 1. Consequences for in vivo digestibility and rate of passage of digesta. Anim
D
245 13. Dugan K, Pond K, Burns J, Fisher D. A comparison of intake and digestibility of Coastal
246 bermudagrass and flaccidgrass hays in horses, in: Proc 13th Equine Nutr Phys Sym
EP
248 14. Eckert JV, Myer RO, Warren LK, Brendemuhl JH. Digestibility and nutrient retention of
C
249 perennial peanut and bermudagrass hays for mature horses. 2010; J Anim Sci 88:2055-
AC
250 61.
251 15. Holland JL, Kronfeld DS, Sklan D, Harris PA. Calculation of fecal kinetics in horses fed hay
253 16. Hussein HS, Vogedes LA, Fernandez GC, Frankeny RL. Effects of cereal grain
255 end-products in the feces and serum of horses consuming alfalfa cubes. J Anim Sci 2004;
256 82:1986-96.
PT
257 17. Lieb S, Mislevy P. Comparative intake and nutrient digestibility of three grass forages:
RI
258 Florakirk and Tifton 85 bermudagrass and florona stargrass to coastal bermudagrass, in:
259 Proc 17th Equine Nutr Phys Symp Lexington, KY, 2001; p. 390
SC
260 18. Miraglia N, Bergero D, Polidori M, Peiretti PG, Ladetto G. The effects of a new fibre-rich
261 concentrate on the digestibility of horse rations. Livest Sci 2006; 100:10-13.
262
U
19. Miyaji M, Ueda K, Kobayashi Y, Hata H, Kondo S. Fiber digestion in various segments of
AN
263 the hindgut of horses fed grass hay or silage. Anim Sci J 2008;79: 339-46.
264 20. Moore-Colyer MJS, Morrow HJ, Longland AC. Mathematical modeling of digesta passage
M
265 rate, mean retention time and in vivo apparent digestibility of two different lengths of hay
D
266 and big-bale grass silage in ponies. Brit J Nutr 2003; 90:109-18.
TE
267 21. Ordakowski-Burk AL, Quinn RW, Shellem TA, Vough LR. Voluntary intake and
268 digestibility of reed canarygrass and timothy hay fed to horses. J Anim Sci.2006;
EP
269 84:3104-09.
270 22. Ott E. Influence of level of feeding on digestion efficiency of the horse, in: Proc 7th Equine
C
272 23. Pagan JD, Jackson SD. Digestibility of long stem alfalfa, pelleted alfalfa or an
273 alfalfa/bermuda straw blend pellet in horses, in: Proc. 12th Equine Nutr Phys Symp
275 24. Pearson RA, Archibald R, Muirhead, RH. The effect of forage quality and level of feeding on
276 digestibility and gastrointestinal transit time of oat straw and alfalfa given to ponies and
278 25. Pearson RA, Archibald RF, Muirhead RH. A comparison of the effect of forage type and
PT
279 level of feeding on the digestibility and gastrointestinal mean retention time of dry
RI
280 forages given to cattle, sheep, ponies and donkeys. British journal of nutrition.
281 2006;95(01):88-98.
SC
282 26. Potts LB, Hinkson JJ, Graham BC, Löest CA, Turner JL. Nitrogen retention and nutrient
283 digestibility in geldings fed grass or alfalfa hay or alfalfa cubes. Journal of Equine
284
U
Veterinary Science. 2009 May 31;29(5):487-8.
AN
285 27. Sponheimer M, Robinson T, Roeder B, Hammer J, Ayliffe L, Passey B, Cerling T, Dearing
286 D, Ehleringer J. Digestion and passage rates of grass hays by llamas, alpacas, goats,
M
287 rabbits, and horses. Small Ruminant Research. 2003 May 31;48(2):149-54.
D
288 28. Staniar WB, Bussard JR, Repard NM, Hall MH, Burk AO. Voluntary intake and digestibility
TE
290 29. Sturgeon LS, Baker LA, Pipkin JL, Haliburton JC, Chirase NK. The digestibility and mineral
EP
291 availability of matua, bermudagrass, and alfalfa hay in mature horses. J Equine Vet Sci
293 30. Thompson K, Jackson S, Baker J. Apparent digestion coefficients and associative effects of
AC
294 varying hay:grain rations fed to horses. Nutr Rep Int 1984; 30: 189-97.
295 31. Clauss M, Schiele K, Ortmann S, Fritz J, Codron D, Hummel J, Kienzle E, The effect of
296 very low food intake on digestive physiology and forage digestibility in horses. J Anim
298 32. Jensen RB, Austbø D, Bach Knudsen KE, Tauson AH. The effect of dietary carbohydrate
299 composition on apparent total tract digestibility, feed mean retention time, nitrogen and
301 33. Miyaji M, Ueda K, Hata H, Kondo S. Effect of grass hay intake on fiber digestion and
PT
302 digesta retention time in the hindgut of horses. J Anim Sci 2014; 92:1574-81.
RI
303 34. Undersander DJ, Mertens DR, Thiex NJ. Forage analyses procedures. National Forage
SC
305 35. Carro MD, López S, Gonález JS, Ovejero FJ. Comparison of laboratory methods for
306 predicting digestibility of hay in sheep. Small Ruminant Research 1994; 14: 9-17.
307
U
AN
308
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
17
309 Table 1
310 Summary of studies used in regression analysis for determining the relationships between forage
Animal Fecal
PT
Study Description of Forage(s) n Gender
Type Collection (d)
Aiken et al.
Bermudagrass hay1 4 Horse Geldings 4
[6]
1
Bergero et al. Natural meadow hay
RI
4(5)a Horse Geldings 6
[7] Sainfoin hay1
Alfalfa hay1
Crozier et al.
Caucasian bluestem1 6 Horse Geldings 5
[8]
Tall fescue hay1
SC
Horses
Cuddeford et Molassed dehydrated alfalfa
8 and Geldings 7
al. [9] hay2
Ponies
Alfalfa hay1,3
U
Cymbaluk
Barley hay3
and
Brome grass hay1 4 Pony Geldings 6
Christensen
AN
Oat hay1,3
[10]
Slough hay1
Alfalfa hay1,3
Altai wildrye hay1
Bromegrass hay1
M
Cymbaluk
6 Pony Geldings 4
[11] Crested wheatgrass hay1
Kentucky bluegrass hay1
Reed canary grass hay1
D
4 Horse Mares 5
[13] Flaccidgrass hay1
Earing et al. Alfalfa hay4
4 Horse Geldings 5
[2] Timothy hay4
Coastal bermudagrass hay1
EP
Eckert et al.
Perennial peanut hay1 6(5)b Horse Geldings 4
[14]
Tifton- 85 bermudagrass hay1
Holland et al.
Alfalfa/orchardgrass mix hay1 4 Horse Geldings 10
[15]
C
Hussein et al.
Alfalfa hay4 4 Horse Geldings 5
[16]
1
AC
Coastal bermudagrass
Lieb and Florakirk bermudagrass hay1
4 Horse Geldings 5
Mislevy[17] Florona stargrass hay1
Tifton-85 bermudagrass hay1
Miraglia et
Meadow hay1 4 Horse Geldings 6
al. [18]
Geldings
Miyaji et al.
Timothy hay1 3 Horse and 4
[19]
Mares
Moore-
Timothy/perennial ryegrass
Colyer et al. 4 Pony Geldings 7
hay1
[20]
ACCEPTED MANUSCRIPT
18
PT
Alfalfa/bermudagrass straw 4 Horse Geldings 5
Jackson [23]
blend3
Pearson et al. Molassed dehydrated alfalfa hay2
4 Pony Geldings 7
[24] Molassed dehydrated oat straw2
RI
Molassed dehydrated alfalfa hay2
Pearson et al. Barley straw2
4 Pony Geldings 7
[25] Mixed grass hay1
Mixed grass hay1
SC
Alfalfa hay1
Potts et al.
Alfalfa cubes4 6 Horse Geldings 4
[26]
Bermudagrass hay1
Sponheimer Smooth bromegrass hay1 Not
4 Horse 5
U
et al. [27] Bermudagrass hay1 reported
Staniar et al.
Teff hay1 6 Horse Mares 3
[28]
AN
Alfalfa hay1
Sturgeon et
Bermudagrass hay1 6 Horse Geldings 4
al. [29]
Matua hay1
Thompson et
Alfalfa hay1
M
4 Horse Geldings 5
al. [30]
1
313 Long stem forage
2
314 Chopped forage
D
3
315 Pelleted forage
4
316 Cubed forage
TE
a
317 One horse added to the study for a period
b
318 One horse removed from the study for a period
319
320
C EP
AC
ACCEPTED MANUSCRIPT
19
321 Table 2
PT
322 Nutrient composition means, ranges, and dry matter digestibility from studies used for model development.
RI
Warm- Warm-
Cool-Season Cool-Season
Cereal Season Season Mixed
Variable Grass Legume All Forages
Forages1
SC
Grain hays Grass Legume
Forages Forages
Forages Forages
n 5 16 17 18 1 8 65
U
NDF % DM 73.8 ± 2.3 66.5 ± 1.4 43.4 ± 1.8 72.0 ± 1.9 46.0 59.3 ± 2.9 61.4 ± 1.7
Range 68.9 – 82.3 59.5 – 76.5 32.8 – 55.0 56.4 – 83.3 - 50.3 – 70.6 32.8 – 83.3
AN
n 6 18 19 18 1 8 70
ADF % DM 45.9 ± 2.6 39.4 ± 0.9 32.8 ± 1.0 37.5 ± 1.0 34.0 34.9 ± 1.6 37.1 ± 0.7
M
Range 34.4 – 52.4 33.2 – 45.9 25.4 – 43.0 28.9 – 43.0 - 30.3 – 40.9 25.4 – 52.4
n 6 18 18 18 1 8 69
CP % DM 5.8 ± 1.3 9.7 ± 0.7 17.4 ± 0.8 9.7 ± 0.7 11.0 13.9 ± 0.9 11.9 ± 0.6
D
Range 2.9 – 11.7 5.9 – 17.1 10.1 – 22.6 6.3 – 16.4 - 10.2 – 16.6 2.9 – 22.6
TE
n 5 16 17 18 1 8 65
HEM3 % DM 25.6 ± 1.3 27.6 ± 1.2 11.2 ± 1.3 34.5 ± 1.4 12.0 24.4 ± 1.6 24.4 ± 1.3
Range 22.8 – 29.9 17.8 – 36.5 3.3 – 22.0 19.6 – 43.3 - 198 – 311 3.3 – 43.3
EP
n 5 16 14 18 1 8 65
4
PER 65.3 ± 1.2 58.6 ± 1.3 75.2 ± 2 52.3 ± 1.3 73.9 58.9 ± 1.3 62.0 ± 1.3
C
Range 62.0 – 68.1 51.2 – 70.2 54.6 – 89.9 45.3 – 65.2 - 53.2 – 65.0 45.3 – 89.9
n 6 18 19 18 1 8 70
AC
DMD 43.0 ± 3.2 48.3 ± 1.3 62.5 ± 1.3 48.2 ± 1.5 65.0 55.7 ± 2.6 52.8 ± 1.1
Range 31.0 – 54.8 38.5 – 58.7 54.4 – 72.8 37.7 – 60.6 - 48.0 – 67.0 31.0 – 72.8
1
323 Mixed forages are diets consisting of two or more forage types.
2
324 Mean ± SE
3
325 HEM = NDF – ADF
4
326 PER = ADF/NDFx100
ACCEPTED MANUSCRIPT
20
327 Table 3
328 Study summary and forage nutrient composition for data sets not included in model development
PT
Study n Gender
(DM basis) Type Collection (d)
Grass hay 1: CP 10.5, ADF 36.0,
Clauss et al. NDF 67.6 Geldings
4 Pony 5
[31] Grass hay 2: CP 5.8, ADF 38.6, and Mares
RI
NDF 69.5
Jensen et al. Timothy hay: CP 10.4, ADF 37.6,
4 Horse Geldings 5
[32] NDF 68.5
SC
Miyaji et al. Timothy hay: CP 11.6, ADF 39.0,
4 Horse Geldings 5
[33] NDF 70.3
330
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
21
331 Table 4
332 Simple linear regression models using forage composition to predict forage digestibility.
Root
Modela P-Value r2 Adj. R2
MSE
PT
DMD = 85.42 – 0.5324×NDF < 0.0001 5.966 0.6017 0.5954
RI
DMD = 93.38 – 1.100×ADF < 0.0001 6.919 0.4643 0.4558
SC
DMD = 67.27 – 0.5938×HEM < 0.0001 7.196 0.4206 0.4114
U
334 calculated values HEM (HEM = NDF – ADF) and PER (PER = ADF/NDF×100).
335
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
22
336 Table 5
337 Multiple linear regression models using forage composition to predict forage digestibility.
Root
Modela P-Value R2 Adj. R2
MSE
DMD = 56.16 + 0.6080×CP –
PT
< 0.0001 5.407 0.6833 0.6677
0.7163×ADF – 0.2560×PER
DMD = 72.85 + 0.6501×CP –
< 0.0001 5.409 0.6830 0.6674
0.5661×ADF – 0.2808×HEM
RI
DMD = 72.85 + 0.6501×CP –
< 0.0001 5.409 0.6830 0.6674
0.2808×NDF – 0.2853×ADFb
DMD = 72.85 + 0.6501×CP –
SC
< 0.0001 5.409 0.6830 0.6674
0.5661×NDF + 0.2853×HEMb
DMD = 80.14 + 0.6931×CP –
< 0.0001 5.444 0.6789 0.6631
0.4412×NDF – 0.1370×PERb
U
DMD = 63.95 + 0.6214×CP –
0.1331×NDFb – 0.5141×ADFb + < 0.0001 5.444 0.6842 0.6631
0.1391×PERb
AN
DMD = 65.81 + 0.7207×CP –
< 0.0001 5.483 0.6690 0.6583
0.3514×NDF
DMD = 66.58 – 0.9966×ADF +
M
0.4084×HEM
DMD = 92.62 – 0.4009×ADF –
< 0.0001 5.791 0.6307 0.6188
0.4084×NDF
DMD = 92.62 – 0.8092×NDF +
C
a
338 Forage composition factors included NDF, ADF, and CP concentrations (DM basis) and
339 calculated values HEM (HEM = NDF – ADF) and PER (PER = ADF/NDF×100).
b
340 Variable not significant (P>0.1)
341
PT
RI
U SC
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
24
342 Fig. 1. Influence of neutral detergent fiber (NDF) concentration on dry matter digestibility
80
70
PT
60
50
RI
DMD (%)
40
SC
30
20
U
10
0
AN
0 20 40 60 80 100
NDF (DM basis)
344
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
25
345 Fig. 2. Influence of acid detergent fiber (ADF) concentration on dry matter digestibility (DMD;
346 DMD = 93.38 – 1.100*ADF; r2 = 0.4643).
80
70
60
PT
50
DMD (%)
RI
r² = 0.4643
30
20
SC
10
0
U
0 10 20 30 40 50 60
AN
NDF (DM basis)
347
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
26
348 Fig. 3. Comparison of regression model predicted and observed digestibility for independent data
349 sets. Observed error bars show SEM and predicted error bars show 95% confidence interval for
PT
Observed CP,NDF CP,ADF,PER CP,ADF,HEM
60
RI
50
Digestibility (%)
40
SC
30
20
U
10
AN
0
Grass Hay 2 [31] Grass Hay 2 [31] Timothy Hay [32] Timothy Hay [33]
351
M
D
TE
C EP
AC