J. Anim Sci. 2007. 85:204-212. doi:10.2527/jas.2005-336
© 2007 American Society of Animal Science
The impact of dietary protein source on observed and predicted metabolizable energy of dry extruded dog foods1
R. M. Yamka*,
K. R. McLeod*,2,
D. L. Harmon*,
H. C. Freetly
and
W. D. Schoenherr
* Department of Animal and Food Sciences, University of Kentucky, Lexington, 40546;
and
USDA, ARS, US Meat Animal Research Center, Clay Center, NE 68933; and
and
Hill’s Pet Nutrition, Topeka, KS 66617
 |
Abstract
|
|---|
Fifty-five observations were used to determine the ME content of 8 foods containing different protein sources. The major protein sources tested included low-oligosaccharide whole soybeans; 2 low-oligosaccharide, low-phytate whole soybeans; 2 conventional soybean meals; low-ash poultry meal; low-oligosaccharide, low-phytate soybean meal; and conventional whole soybeans. The ME content of all foods ranged from 3,463 to 4,233 kcal/kg of DM. The first objective was to utilize the observed ME data and test the accuracy of the modified Atwater equation. In this study, the modified Atwater equation generally underpredicted ME compared with the observed ME (residual mean = 247 kcal/kg). The second objective was to use individual data to develop an equation, based on the chemical composition of the food, to predict the ME content of the foods. A multivariate regression analysis was used to predict ME content based on chemical composition. Five models were fitted to the data. Model 1 included CP, ether extract (EE), and crude fiber (CF). Because the foods varied in protein sources, and the ratio of total AA (TAA) to non-AA (NAA) CP ranged from 3.5:1 to 14.4:1, it was hypothesized that accounting for the proportion of TAA and NAA in CP would improve the fit of the model. Therefore, model 2 included TAA, NAA, EE, and CF. Defining CP in terms of TAA and NAA improved the r2 of the model from 0.46 to 0.79. Subsequently, models 3, 4, and 5 replaced the CF term with ADF, NDF, and hemicellulose (HEM). Model 3 included TAA, NAA, EE, and NDF. Model 4 included TAA, NAA, EE, ADF, and HEM. Model 5 included TAA, NAA, EE, and HEM. Defining dietary fiber in terms of HEM improved the r2 of model 2 from 0.79 to 0.81. Residual analysis suggested that replacing the CF term with HEM (model 5) improved the prediction of ME content. In contrast, defining fiber in terms of NDF (model 3) did not result in an improvement over model 2, whereas the ADF term (model 4) did not (P > 0.34) contribute to the overall model. Fractionating CP into TAA and NAA components further defined the chemical composition of the food. These data suggest that defining protein composition improves the accuracy of predicting the ME content of dog foods.
Key Words: amino acid • crude protein • dog • fiber • metabolizable energy
|
INTRODUCTION
|
|---|
A key function of dietary intake is to provide energy. In contrast to individual or specific nutrients, there is an acceptable, low tolerance range for errors in determining dietary energy values for dogs because under- and overestimates lead to excessive weight loss or gain. Nevertheless, because of the laborious nature of determining the actual ME content of dog foods, calculated ME values are used extensively for diet formulation and food labeling.
The current Association of American Feed Control Officials (AAFCO, 2004
) and National Research Council (NRC, 1985) guidelines recommend a predictive equation that is largely based on fixed energy values and digestibility coefficients for dietary components (i.e., CP, crude fat, and carbohydrate) for estimating ME content of dog foods. Although this predictive equation has provided adequate precision for foods containing traditional ingredients, it may be inadequate for currently available commercial foods due in part to the use of new food additives and diverse ingredients.
Recent research has focused on the fiber portion of the food to attempt to find a more accurate method of determining the ME (Earle et al., 1998; Kienzle, 2002). However, the crude fiber (CF) content of most commercial dog foods accounts for only approximately 3 to 5% of the food. Therefore, it seems reasonable to focus on other components of the food, such as CP (20 to 30% of the dietary DM), which would contribute more to total energy content and, thus, may have a greater impact on the prediction of ME.
The CP fraction represents numerous compounds that can broadly be classified as total AA (TAA) and non-AA compounds (NAA; e.g., nucleic acids, amines, amides, etc.). Digestibility (Giesecke et al., 1982; Yamka et al., 2005a) and relative contribution to urinary N energy losses (Blaxter, 1989) differ among these compounds. Thus, variation in the TAA:NAA ratio among dietary sources, including plant and animal CP (NRC, 1996), can introduce inherent errors in estimating dietary ME when CP is used in the prediction equation. We hypothesized that dividing CP into TAA and NAA would improve the precision of predicting dietary ME.
Therefore, the objective of this study was to determine the actual ME in foods containing different protein sources and to test if the prediction of dietary ME is improved by defining CP in terms of TAA and NAA.
|
MATERIALS AND METHODS
|
|---|
Animal Care
The dogs were located in the Division of Laboratory Animal Research Facility at the University of Kentucky and were cared for in accordance with Institutional Animal Care and Use Committee protocols. Energy balance was determined (55 observations) using 16 adult, spayed, female mongrel dogs (19.5 ± 0.3 kg of BW). Environmental conditions, descriptions of the individual kennels, and protocols for total urine and fecal collection have been previously described (Yamka et al., 2005b). Cleaning was done twice daily, after feeding time. During total fecal and urine collection, the dogs were confined to their primary living quarters to ensure that all urine and feces were collected. Dogs were exercised daily, and human interaction included, but was not limited to, cage play (toys), grooming, and other human-dog interactions (e.g., petting). Dogs had ad libitum access to water throughout the experiment.
Experimental Design
Data were generated from 2 replicated, 4 x 4, Latin square studies. In total, 8 protein sources were formulated into 8 complete foods (Tables 1 and 2) in accordance with the AAFCO (2004) guidelines and used in energy balance trials (55 observations) to determine the ME value of foods containing each of the protein sources. The major protein sources included low-oligosaccharide whole soybeans (LO; ME = 3,985 kcal/kg); low-oligosaccharide, low-phytate whole soybeans (LLB1; ME = 3,915 kcal/kg and LLB2; ME = 3,614 kcal/kg); 2 conventional soybean meals (SBM1; ME = 4,066 kcal/kg and SBM2; ME = 3,737 kcal/kg); low-ash poultry meal (PM; ME = 4,233 kcal/kg); low-oligosaccharide, low-phytate soybean meal (ME = 3,890 kcal/kg); and conventional whole soybeans (WSB; ME = 3,463 kcal/kg).
View this table:
[in this window]
[in a new window]
|
Table 1. Ingredients of dog foods containing LO, LLB1, PM, SBM1, LLM, SBM2, LLB2, and WSB as dietary protein sources (as-fed basis)1
|
|
View this table:
[in this window]
[in a new window]
|
Table 2. Proximate analysis of dog foods containing LO, LLB1, PM, SBM1, LLM, SBM2, LLB2, and WSB as dietary protein sources (DM basis)1
|
|
At the beginning of each trial, intake was adjusted to supply the proper amount of energy required for maintenance (1,258 kcal of ME/d). Maintenance energy requirements (kcal/d) were calculated as 1.6 times the resting energy expenditure, where resting energy expenditure = 70 kcal/kg of BW0.75, and 1.6 represents the estimated increase in heat production above that of a resting state for a neutered dog (Gross et al., 2000).
Each balance trial consisted of a 6-d period for adaptation to the food, followed by a 5-d period of total fecal and urine collection. During the collection period, feces were weighed, frozen (–20°C) daily, and at the end of the period, a composite sample from each dog was generated. Similarly, urine was collected into an acidified vessel (5 mL, 6 M phosphoric acid), and a composite for each dog was created by pooling daily urine samples (kept at 4°C) throughout the collection period, after which an aliquot from each was frozen at –20°C. A minimum of 8 d were allowed between measurements of energy balance for each food.
Sample Analysis.
Feed and fecal samples were lyophilized (Dura-Dry MP Freeze-Drier, FTS Systems, Stone Ridge, NY) and ground through a 0.5-mm screen (Cyclotec 1093 Sample Mill, Tecator, Hoganas, Sweden). The CP content of the samples was obtained using a Leco CN2000 N analyzer (Leco Corp., St. Joseph, MI). Ether extract (EE), CF, NDF, ADF, and ash of feed samples (Table 1
) were determined according to methods 954.02, 962.09, 973.18, and 942.05 of the AOAC (1995). The hemicellulose (HEM) content was calculated as the difference between NDF and ADF. Nitrogen-free extract (NFE) was calculated as 100 minus the sum of the percentages of CP, EE, CF, moisture, and ash (AAFCO, 2004).
Food samples were analyzed for AA concentrations according to methods 988.15 (sulfur and regular) and 994.12 (Trp) of the AOAC (1995). The resulting solutions were derivatized with 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate (AccQ·Tag Chemistry Package, Millipore Waters, Milford, MA), and the AA concentration was determined by reverse phase, liquid chromatography (Alliance 2695 Separation Module, Waters) as described by Liu et al. (1995). Total AA were determined to be the sum of all AA. Non-AA CP was the difference between CP and TAA. The GE of feed, fecal, and urine samples (Table 2) were determined using a 1261 Isoperibol Parr Bomb calorimeter (Parr Instrument Co., Moline, IL). Metabolizable energy was calculated as DE minus urinary energy loss. Energy lost as methane was not accounted for in the calculation of ME because this loss is considered to be negligible in dogs (McKay and Eastwood, 1984; NRC, 1985).
Statistical Analyses
Data were analyzed using the GLM procedure (SAS Inst. Inc., Cary, NC). Mean DE and ME values for each food and mean and SD across foods were generated from the observed data (Table 3). The data set for analysis included only data from dogs that completely consumed their allocated food (n = 55). Accordingly, data from 3 dogs were excluded from the data set because of insufficient food intake. Residual analysis of the predicted, modified Atwater ME for foods was performed to determine if a bias existed in ME prediction. Data were analyzed to determine a more accurate fit for the prediction of ME. This was accomplished by using the REG procedure of SAS to determine if the addition of CP, AAN, NAA, EE, CF, NDF, ADF, and HEM to the model contributed to a more accurate prediction of ME. Five models were evaluated and included the following input variables: 1) CP, EE, and CF; 2) TAA, NAA, EE, and CF; 3) TAA, NAA, EE, and NDF; 4) TAA, NAA, EE, ADF, and HEM; and 5) TAA, NAA, EE, and HEM.
View this table:
[in this window]
[in a new window]
|
Table 3. Observed energy values of dog foods containing LO, LLB1, PM, SBM1, LLM, SBM2, LLB2, and WSB as dietary protein sources1
|
|
|
RESULTS
|
|---|
After the conclusion of the experiments, the energy balance data were used to challenge the modified At-water equation. The modified Atwater equation, ME (kcal/kg) = (3.5 x CP) + (8.5 x EE) + (3.5 x NFE), is currently used in the pet food industry to predict the ME content of dry extruded dog foods (AAFCO, 2004; NRC, 1985). Residual analysis of the predicted modified Atwater ME for foods in this analysis is shown in Figure 1. Across all foods the modified Atwater equation generally underestimated the observed ME (residual mean = 247 kcal/kg). Mean residual ME was greatest for foods containing PM (546 kcal/kg of DM) as the primary source of dietary protein and least for those containing WSB (42 kcal/kg of DM) or SBM2 (–13 kcal/kg of DM).
Because the Atwater equation failed to accurately predict the ME of the foods tested, 5 models were developed to fit the data (Table 4). Model 1 included CP, EE, and CF. In model 2, CP was replaced with its constituent terms, TAA and NAA, which resulted in a model that contained TAA, NAA, EE, and CF. Model 1 had an r2 of 0.46; however, when TAA and NAA were substituted for CP, the r2 of the model improved to 0.79. Residual analysis indicated replacing CP with TAA and NAA (Figure 2) in model 1 improved prediction of ME, and no residual bias (residual mean = 0 kcal/kg; P = 1.0 and slope = 0; P = 1.0) was observed. Models 3, 4, and 5 further defined the food by separating dietary fiber into NDF, ADF and HEM, and HEM, respectively. In model 3, CF was replaced by NDF. No improvement in the prediction of ME was observed (r2 = 0.72). Model 4 replaced the single term CF with ADF and HEM. The ADF portion did not contribute (P = 0.34) to the prediction of ME. As a result, this model was excluded from Table 4. When HEM was substituted for CF (model 5), the r2 of the model increased from 0.79 to 0.81. Residual analysis suggested that by replacing the CF term in model 2 with HEM in model 5 (Figure 3), there was a slight improvement in prediction of ME content with no residual bias (residual mean = 0 kcal/kg; P = 1.0 and slope = 0; P = 1.0).
View this table:
[in this window]
[in a new window]
|
Table 4. Regression coefficients for the prediction of ME content (kcal/kg) of dog foods fed at maintenance based on dietary chemical constituents (% of total DM)1
|
|
Because DE comprises the majority of ME, the same models were used to predict DE (Table 5). Model 1 had an r2 of 0.48; however, when TAA and NAA were substituted for CP, the r2 of the model improved to 0.81. Models 3, 4, and 5 further defined the food again by separating the CF into NDF, ADF and HEM, and HEM, respectively. Model 3 showed no improvement in the prediction of ME (r2 = 0.72). Similar to the ME prediction, the ADF portion (model 4) did not contribute to the prediction of DE (P = 0.42). As a result, this model was excluded from Table 5. Model 5 showed a slight increase in the prediction of DE (r2 = 0.83).
View this table:
[in this window]
[in a new window]
|
Table 5. Regression coefficients for the prediction of DE content (kcal/kg) of dog foods fed at maintenance based on dietary chemical constituents (% of total DM)1
|
|
|
DISCUSSION
|
|---|
Energy balance measurements, in which 8 different foods were fed (Table 1) for a total of 55 observations, were used to provide a data set to challenge the current AAFCO (2004) and NRC (1985) factorial prediction equation (i.e., modified Atwater) for dietary ME and subsequently to develop an alternative regression equation based on the chemical composition of dry extruded dog foods. Experimental foods were formulated to provide equal amounts of energy and CP; however, the primary source of protein varied between foods and included animal and plant sources. Across all foods, the observed DE and ME averaged 4,043 and 3,863 kcal/kg of DM, respectively. However, when foods were compared, PM had the highest overall observed DE and ME (4,417 and 4,233 kcal/kg of DM, respectively), and WSB had the lowest (3,624 and 3,463 kcal/kg of DM, respectively). These results were expected because animal protein sources are generally considered to be more digestible and available to the dog than plant protein sources (Murray et al., 1997).
Using the current data set, a plot of the residual ME values (i.e., observed ME minus that derived from the modified Atwater equation), shows that the modified Atwater equation generally underpredicted ME across foods (Figure 1). However, further examination of mean residual ME values for each food reveals a bias in the equation. Case-in-point, mean residual ME values for foods containing WSB and SBM2 were 42 and 13 kcal/kg of DM, respectively, indicating that the predicted values fit the observed data reasonably well. In contrast, mean residual ME values ranged from 156 to 546 kcal/kg of DM for foods containing PM, nonconventional soybeans (i.e., LO, LLB1, LLB2, and LLM), and SBM1, representing an underprediction of observed ME by the Atwater equation. In the case of the PM-containing food, the Atwater equation underpredicted the observed ME (3,667 kcal/kg of DM) by 14.9%. The failure of the Atwater equation to accurately predict ME content of the foods presents practical problems that can result in underfed or overfed dogs. In particular, the underestimation of the ME content could result in obesity and obesity-related disorders (i.e., diabetes and arthritis) simply because the consumer is following recommended feeding instructions for what is believed to be a lower calorie food. Conversely, overestimation would inadequately predict ME for growth, lactation, or posttrauma recovery in dogs.
Test foods used in the current data set varied in ingredients (Table 1); however, the primary variant was protein source. Thus, with the exception of SBM2 and WSB, the inability of the Atwater equation to accurately predict ME in these foods may be due to the inclusion of protein sources that deviate from those in the foods from which the modified Atwater equation was derived. The Atwater equation assumes constant total-tract apparent digestibility coefficients for the CP, EE, and nonstructural carbohydrate (NFE) fractions of the food. In the current data set, apparent total-tract digestibility of the EE and NFE fractions was not measured. However, CP digestibility of these same test foods averaged 0.82 and ranged from 0.77 (LLB2) to 0.86 (PM; Yamka et al., 2005a,b). Considering that the modified Atwater equation (NRC, 1985) assumes a CP digestibility of 0.80, the underprediction of ME in some of the test foods in the current data set may be, at least in part, a function of underestimating CP digestibility. However, CP digestibility for the test foods in which ME was predicted reasonably well by the Atwater equation, ranged from 0.77 (WSB) to 0.85 (SBM2). Likewise, the Atwater equation underpredicted the ME content of the test food containing LLB2 as the primary protein source despite CP digestibility (0.77) being substantially lower than the assumed digestibility. Therefore, the failure of the Atwater equation to accurately predict ME content in the current test foods does not seem to be simply a function of CP digestibility.
Fractioning CP into TAA and NAA further defines the chemical composition of the food. The significance of this fractionation is demonstrated by comparing models 1 and 2 (Table 4). Dividing CP into its constituent components improved the r2 of the ME prediction equation from 0.46 (model 1) to 0.79 (model 2). In model 1, where CP is not fractionated, the coefficient for CP is negative. However, as observed in model 2, the actual coefficient for TAA is positive, whereas that for NAA is negative. The relative contribution of components to the model was determined using standard regression estimates. In model 2, the relative contributions of TAA and NAA were 0.04 and 0.32, respectively. This indicates that the contribution of TAA to the model is relatively minor, whereas that from NAA model is greater or slightly less than that of CF (0.23) and EE (0.42). The disparity in the relative contributions of TAA and NAA to the model may in part be explained by the observed variation in digestibility within these fractions. In the test foods used in the current data set, ileal digestibility of the TAA fraction varied from 0.69 to 0.83, whereas that for NAA varied from 0.08 to 0.78 (Yamka et al., 2005a,b).
Because methane production is considered to be negligible in dogs (McKay and Eastwood, 1984), ME can be defined in terms of DE and urinary energy losses. Accordingly, improvement in the ME model by partitioning CP into TAA and NAA is reflective of one or both of these factors. The increase in r2 observed for model 2 compared with model 1 is the same for the ME and DE (Table 5) prediction models. This suggests that the improved accuracy in ME prediction resulting from the fractionation of CP into TAA and NAA is attributable to digestibility. In the current data set, ileal digestibility of CP averaged 0.75 across foods; however, the average digestibility of the TAA (0.78) and NAA (0.50) fractions was substantially different (Yamka et al., 2005a,b). Although the NAA fraction was calculated in the current study as the difference between CP and TAA, the low digestibility of this fraction is consistent with earlier observations that have shown nucleic acids, a primary component of the NAA fraction, have a low digestibility in dogs (Giesecke et al., 1982). Moreover, the difference in coefficients for TAA and NAA between the DE and ME models reflects energy lost in the conversion of DE to ME; in this case, representing primarily urinary energy loss. Considering model 2, the differences in coefficients between DE and ME were –12.3 and –9.9 for TAA and NAA, respectively. This indicates that a greater fraction of the total urinary energy was derived from TAA than NAA energy because of the differences in dietary supply (16.9 and 3.0% DM, respectively). However, the similar differences in coefficients between DE and ME for TAA and NAA indicate that the proportion of dietary TAA and NAA energy lost in the urine is largely the same for both fractions. Thus, the improvement in ME prediction associated with fractionating CP into TAA and NAA is not attributable to urinary loss but rather to differences in digestibility between the 2 fractions.
Many of the recent studies related to food energy have focused on the fiber portion of the food because it has been shown to have a negative impact on DE and ME values of pet foods (Kienzle et al., 1998; Laflamme, 2001; Kienzle, 2002). Earle et al. (1998) found a negative relationship between GE digestibility and dietary fiber content and described the relationship between DE and dietary CF content using the following regression equation: apparent digestible energy (%) = 94.32 – (1.43 x CF as a percentage of DM). When this equation was applied to the current data set, residual DE ranged from 43 to –244 kcal/kg of DM (data not shown). Consistent with the studies cited above, the negative coefficients for CF in the current DE and ME prediction models (1 and 2) reflect a negative relationship between CF and DE and ME. When dietary fiber was described in terms of NDF (i.e., HEM, cellulose, and lignin), ADF (i.e., cellulose and lignin), and HEM for DE and ME prediction, positive regression coefficients were generated for NDF and HEM, whereas ADF did not contribute significantly to the overall models. Given the narrow range in fiber content across test foods used in the current data set, the positive coefficients for NDF and HEM, in contrast to those for CF, reflect greater digestibility of these fiber fractions compared with ADF (Fahey et al., 1990a,b). For DE and ME prediction, substitution of CF (Model 2) with NDF (Model 3) decreased the r2 from 0.79 to 0.72; thus accounting for less variation than that obtained with CF in the model. Because the NDF content of test foods used in the current data set varied a maximum of 2 g, the reduction in r2 is inherent to the data set used. Defining the fiber fraction solely in terms of HEM (model 5) resulted in a slight increase in r2 (0.79 to 0.81) with a relative contribution of 0.24 to the model. Perhaps a greater amount of variation could be accounted for by the inclusion of all fermentable fiber (i.e., soluble and HEM) in the model. Although these fiber levels represent typical levels in most commercial dog foods, the range in the fiber fractions was minimal. Therefore, the current data set is not robust enough to adequately test the utility of fractionating fiber in order to improve the accuracy of DE and ME prediction.
The modified Atwater equation has been used as an industry standard for estimating ME content of extruded dry dog food. However, the use of this equation can result in improper feeding and formulation of foods that are meant to be low-calorie because actual ME intake is much higher than predicted. Accounting for the proportions of TAA and NAA in CP dramatically increased the r2 (0.46 to 0.79) of the model, reflecting an improvement in the accuracy of predicting the ME content of dry extruded dog foods containing different protein sources. The fractionation of CP accounts for compositional differences instead of relying on total CP when formulating foods with different protein sources. Because many dog foods are high in CP, containing greater than 30.0% and are composed of a mixture of protein sources, fractionation of the CP into TAA and NAA should be considered when predicting ME content. Use of the model 2 regression equation, or incorporation of TAA and NAA fractions into existing prediction models, would not be difficult because TAA content of diets are already in use in the pet food industry in order to formulate diets to meet AA requirements.
Most typical dog foods contain minimal amounts of CF, with the exception of light foods, ranging from 3.0 to 5.0% CF of the total food. Accounting for the proportions of HEM in the food marginally improved ME prediction when compared with CF; however, more research focusing on diverse fiber sources, intake, and perhaps fermentable fiber content is necessary to determine which components may significantly affect ME prediction.
|
IMPLICATIONS
|
|---|
The modified Atwater equation is used extensively to estimate the metabolizable energy value of commercial dog foods. However, because the Atwater equation is dependent upon fixed digestibility coefficients for macronutrients, it is susceptible to prediction errors. The current study demonstrates that the Atwater equation introduces a bias in the prediction of the metabolizable energy content of dry extruded foods containing varying sources of protein. Multivariate regression analysis, based on dietary chemical composition, shows that accuracy of prediction improved by fractionating dietary crude protein into amino acid and non-amino acid components.
|
Footnotes
|
|---|
1 Published as publication no. 03-07-087 of the Kentucky Agric. Exp. Station. 
2 Corresponding author: kmcleod{at}uky.edu
Received for publication June 28, 2005.
Accepted for publication August 9, 2006.
|
LITERATURE CITED
|
|---|
AAFCO. 2004. Association of American Feed Control Officials: Official Publication. The Association, Atlanta, GA.
AOAC. 1995. Official Methods of Analysis. Assoc. Off. Anal. Chem., Washington, DC.
Blaxter, K. L. 1989. Energy Metabolism in Animals and Man. Cambridge Univ. Press, New York, NY.
Earle, K. E., E. Kienzle, B. Optiz, P. M. Smith, and I. E. Maskell. 1998. Fiber affects digestibility of organic matter and energy in pet foods. J. Nutr. 128:2798S–2800S.[Free Full Text]
Fahey, G. C., Jr., N. R. Merchen, J. E. Corbin, A. K. Hamilton, K. A. Serbe, S. M. Lewis, and D. A. Hirakawa. 1990a. Dietary fiber for dogs: I. Effects of graded levels of dietary beet pulp on nutrient intake, digestibility, metabolizable energy, and digesta mean retention time. J. Anim. Sci. 68:4221–4228.[Abstract]
Fahey, G. C., Jr., N. R. Merchen, J. E. Corbin, A. K. Hamilton, K. A. Serbe, S. M. Lewis, and D. A. Hirakawa. 1990b. Dietary fiber for dogs: II. Iso-total dietary fiber (TDF) additions of divergent fiber sources to dog diets and their effects on nutrient intake, digestibility, metabolizable energy, and digesta mean retention time. J. Anim. Sci. 68:4229–4235.[Abstract]
Giesecke, D., S. Gaebler, and W. Tiemeyer. 1982. Purine availability and metabolism in dogs fed single-cell protein or RNA. J. Nutr. 112:1822–1826.[Abstract/Free Full Text]
Gross, K. L., K. J. Wedekind, C. S. Cowell, W. D. Schoenherr, D. E. Jewell, S. C. Zicker, J. Debraekeleer, and R. A. Frey. 2000. Nutrients. Pages 21–107 in Small Animal Clinical Nutrition. M. S. Hand, C. D. Thatcher, R. L. Remillard, and P. Roudebush, ed. Walsworth Publishing Co., Marceline, MO.
Kienzle, E. 2002. Further developments in the prediction of metabolizable energy (ME) in pet food. J. Nutr. 132:1796S–1798S.[Abstract/Free Full Text]
Kienzle, E., B. Optiz, K. E. Earle, P. M. Smith, I. E. Maskell, and C. Iben. 1998. An improved method for the estimation of energy in pet foods. J. Nutr. 128:2806S–2808S.[Free Full Text]
Laflamme, D. P. 2001. Determining metabolizable energy content in commercial pet foods. J. Anim. Physiol. Anim. Nutr. (Berl.) 85:222–230.[Medline]
Liu, H. J., B. Y. Chang, H. W. Yan, F. H. Yu, and X. X. Liu. 1995. Determination of amino acids in food and feed by derivatization with 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate and reverse-phase liquid chromatographic separation. J. AOAC Int. 78:736–744.
McKay, L. F., and M. A. Eastwood. 1984. A comparison of bacterial fermentation end products in carnivores and primates including man. Proc. Nutr. Soc. 43:35A.[CrossRef]
Murray, S. M., A. R. Patil, G. C. Fahey, Jr., N. R. Merchen, and D. M. Hughes. 1997. Raw and rendered animal by-products as ingredients in dog diets. J. Anim. Sci. 75:2497–2505.[Abstract/Free Full Text]
NRC. 1985. Pages 2–41 in Nutrient Requirements of the Dog. Natl. Acad. Press, Washington, DC.
NRC. 1996. Pages 191–213 in Nutrient Requirements of Beef Cattle. 7th rev. ed. Natl. Acad. Press, Washington, DC.
Yamka, R. M., B. M. Hetzler, and D. L. Harmon. 2005a. Evaluation of low-oligosaccharide low-phytate whole soybeans and soybean meal in canine foods. J. Anim. Sci. 83:393–399.[Abstract/Free Full Text]
Yamka, R. M., S. E. Kitts, and D. L. Harmon. 2005b. Evaluation of low-oligosaccharide and low oligosaccharide low-phytate whole soya beans in canine foods. Anim. Feed Sci. Technol. 120:79–91.[CrossRef]
Top
Abstract
INTRODUCTION
Low-oligosaccharide whole soybeans (LO; n = 6); 
low-oligosaccharide, low-phytate whole soybeans (LLB1; n = 6); 
low-ash poultry meal (PM; n = 6); 
conventional soybean meal (SBM1; n = 6); x low-oligosaccharide, low-phytate soybean meal (LLM; n = 8); • conventional soybean meal (SBM2; n = 8); + low-oligosaccharide, low-phytate whole soybeans (LLB2; n = 7); and –conventional whole soybeans (WSB; n = 8).
