INTRODUCTION
Meat and bone meal (MBM) is the dried and rendered product of mammalian tissue and consists of animal offal, bones, blood, heads, lean tissues, and fat and is generally regarded as a good source of CP and AA for swine (Traylor et al., 2005) and chickens (Huang et al., 2005). However, there has been less emphasis on energy value of MBM for swine.
Batterham et al. (1980) reported a range of 2,247 to 3,322 kcal/kg for DE of 14 MBM for growing pigs. Shi and Noblet (1993) determined ME in sows and growing pigs; they reported the ME value of 3,009 and 2,175 kcal/kg of DM for sows and growing pigs, respectively. Recently, Adedokun and Adeola (2005) in a digestion trial determined the apparent ME (AME) and apparent nitrogencorrected ME (AME_{n}) values of 12 MBM samples and reported a range of 1,569 to 3,308 kcal/kg of DM and 1,474 to 3,361 kcal/kg of DM, respectively. Dolz and De Blas (1992) observed that in chickens the differences in energy contents of MBM could be attributed to GE content and that GE tended to increase with an increase in the content of fat in the meal. Adedokun and Adeola (2005) noted that the variation in AME and AME_{n} were not related to any of the major chemical components, but that interactions among the components may be a factor.
Differences in species of origin, quantity of bones, and other factors produce variability in the proximate composition and energy value of MBM. An assay of MBM from different sources that are different in composition may provide a prediction equation that is robust enough to be used in the prediction of energy value of MBM. Therefore, the objective of this study was to determine the AME and AME_{n} of 21 MBM samples and establish equations for predicting energy values of MBM for swine.
MATERIALS AND METHODS
All animal experimentation procedures were approved by the Purdue Animal Care and Use Committee.
Animal and Diets
A total of 198 crossbred (Hampshire × Duroc × Yorkshire × Landrace) barrows with an average initial BW of 32.0 kg were used for this study. The barrows were allocated to 22 dietary treatments in a randomized complete block design with each treatment replicated 9 times. The barrows were housed in stainlesssteel metabolism crates that allowed for total but separate collection of feces and urine using a 5d adjustment period followed by 5 d of feeding the experimental diets according to the procedure of Adeola and Bajjalieh (1997). The 22 experimental diets included a cornsoybean meal reference diet (RD); the others were test diets (TD) in which each of the 21 MBM replaced part of corn and soybean meal in the RD such that the ratio of corn to soybean meal was the same in the RD and TD. Maintaining a constant ratio for the feedstuffs supplying energy in reference and test diets is a basic requirement to enable the use of the difference method for determining the ME content of a test ingredient. The ingredient compositions of RD and TD are presented in Table 1.
Chemical Analyses
Dried fecal, orts, and feed samples were ground to pass through 0.5mm screen using a mill grinder (Retsch ZM 100, GmbH & Co. K.C., Haan, Germany). Urine samples for each barrow were thawed and thoroughly mixed, after which two 800mL subsamples were filtered 3 times using glass wool and then dried in a forcedair oven. The dried urine was stored at −20°C before analysis. For DM determination, samples were dried at 100°C in a drying oven (Precision Scientific Co., Chicago, IL) for 24 h (AOAC, 2000). Gross energy was determined in a bomb calorimeter (Parr 1261 bomb calorimeter, Parr Instruments Co., Moline, IL) using benzoic acid as a calibration standard. Nitrogen was determined by combustion method (Leco Model 2000 CHN analyzer, Leco Corp., St. Joseph, MI) using EDTA as a calibration standard.
The proximate and AA composition of the MBM samples were determined at the University of Missouri Experimental Station, Columbia. For AA analysis, MBM samples were hydrolyzed in 6 N HCl for 24 h at 110°C under N atmosphere. For Met and Cys, performic acid oxidation was carried out before acid hydrolysis. The AA in the hydrolyzate were determined by HPLC after postcolumn derivatization [AOAC, 2000, method 982.30 E (a,b,c)].
Calculations
For each of the 22 diets, energy digestibility (ED, %) and energy metabolizability (EM, %) were calculated. Consequently, DE (kcal/kg) and ME (kcal/kg) of the diets were calculated by multiplying ED and ME (%), respectively, with GE (kcal/kg) of the diets.
To calculate the DE, AME, or AME_{n} of the test ingredients (MBM), the following series of equations (applicable to both DE and AME) were used:where EM_{TD} is the energy metabolizability (%) for the TD; EM_{RD} is the energy metabolizability (%) for the RD; EM_{TI} is the unknown energy metabolizability for the test ingredient (TI); FC_{RD/TD} is the fractional contribution of GE (kcal/kg) in the RD to TD in the diet; and FC_{TI/TD} is the fractional contribution of GE (kcal/kg) in the TI to the TD.
By definition, for energy or any other nutrient: FC_{TI/TD} + FC_{RD/TD} = 1. Therefore, FC_{RD/TD} = 1 − FC_{TI/TD}. [2] Hence, substituting [2] in [1] gives
Solving for the only unknown, EM_{TI}, gives
The AME (kcal/kg) of TI was calculated as follows:where AME_{TI} and GE_{TI} are in kcal/kg. The AME corrected for retained nitrogen was calculated using a caloric value of 7.45 kcal/g of N (Harris et al., 1972).
Statistical Analyses
The data on DE, AME, and AME_{n} for the diets and MBM samples were analyzed using the GLM procedure (SAS Inst. Inc., Cary, NC). For the statistical analyses described below, the means for AME and AME_{n} for the 21 MBM samples were used along with their respective proximate analysis. To study the relationship among the proximate compositions and DE, AME, and AME_{n} of the MBM, the CORR procedure of SAS was used, whereas the INSIGHT procedure of SAS as well as regression diagnostics (DFFITS, DFBETA, and Cook’s distance) were used for the multiple regression analysis and for identifying possible influential points and outlying observations. The measure of the influence of a particular case on each of the regression coefficients is denoted as DFBETA; a data point is influential if DFBETA >1. For multicolinearity diagnostics, tolerance and variance inflation factors (VIF) were used. Excessive colinearity is indicated by tolerance >10 or VIF <0.1. It was assumed a priori that there is colinearity among ash, P, and Ca; therefore, these were not considered in colinearity diagnostics.
To develop the optimum regression model, model selection criteria such as Mallows’ Cp (which compares full with reduced models for error sums of squares) and Akaike information criterion (AIC) were used along with stepwiseregression analysis, forward selection, backward elimination, and maximum R^{2} improvement. The results obtained in these steps were compared with other model selection criteria listed above. The best 2 models for each subset of combination of explanatory variables were tested, and the optimum subset was chosen. The criterion Mallows’ Cp was used for model selection such that the effect of multicolinearity on the fit of the regression model could be avoided. Multicolinearity may result from having many variables in the regression model and may produce undesirable overfit of the model. Mallows’ Cp assigns values for each model, and the best model is the one that is approximately equal to the number of parameters (variables) in the model. Within each variable combination subsets, however, the best and unbiased model is the one with least Cp value or with Cp ≈p (where p is the number of parameters in the model). The full model is not considered in using Cp as a selection criterion because for the full model Cp is always equal to p. For the selection of the final model to be used in describing AME and AME_{n} of the MBM, AIC was used as the selection criterion. This model selection criterion is based on the loglikelihood of the model and penalizes for addition of parameters (model complexity); thus, it is capable of selecting a model that fits well but also has the minimum number of variables. Smaller AIC values indicate better fit.
RESULTS
The chemical composition of the 21 MBM samples is shown in Table 2. The GE content ranged from 3,895 to 5,193 with an average of 4,601 kcal/kg of DM. The nutrient contents were, on average, 589, 113, 239, 38, and 77 g/kg of DM for CP, fat, ash, P, and Ca, respectively. In addition, the total AA content of the MBM ranged from 464 to 603 g/kg of DM with an average of 550 g/kg of DM (Table 3).
Generally, energy utilization responses, except AME_{n}, were least (P < 0.05) in the RD compared with the TD (data not shown). On average, DE, AME, and AME_{n} values for the diets were 4,045, 3,940, and 3,733 kcal/kg, respectively. Nitrogen and DM digestibility were on average 86.2 and 83.5%, respectively. Digestible energy, AME, and AME_{n} of the MBM are presented in Table 4. On average, DE, AME, and AME_{n} were 3,402, 3,069, and 2,963 kcal/kg, respectively.
Using DFBETA criterion, MBM 4 was an influential point because of the values of the DFBETA for CP and fat; hence, the data from this MBM were removed from further analyses. Correlation coefficients of the proximate fractions with GE, DE, AME, and AME_{n} of the MBM are presented in Table 5. None of the correlations of the proximate fractions with the energy utilization responses were >0.50. Only fat was positively correlated with DE, AME, and AME_{n}; r values were 0.41, 0.42, 0.40, and 0.44 for AME, AME_{n}, DE, and GE, respectively. Calcium, ash, P, and CP were all negatively correlated with all the measures of energy utilization except for CP, which was positively correlated with GE. The greatest absolute rvalues were observed for the correlations between proximate fractions and GE compared with the values for correlation of the proximate fractions with DE, AME, and AME_{n} of the MBM. In Table 6, the correlations of the energy measures with the ratios of proximate fractions to each other are presented. Except for CP:fat and GE:fat, all the ratios of proximate fractions were positively correlated with all the energy utilization measures. The greatest r of 0.91 was observed for correlation of CP + fat:ash with GE. In addition, GE:fat was positively correlated with GE of the MBM.
Table 7 shows the variables used for the model development. The variables used in the model building process were fat, ash, CP, Ca, P, and GE. Regression coefficient and Cp were the criteria used for model selection; the best 2 models within each subset of the variables from 1 to 5variable models are presented. The variables entered in the models were the same for AME and AME_{n}. For AME and AME_{n}, fat was the best predictor in the 1variable model because of its greater R^{2} value and lesser Cp compared with ash, the next best predictor in the 1variable model. The best 2variable models incorporated CP and ash, whereas the best 3variable model used GE, P, and ash. The best 4variable model included CP with the variables already in the 3variable model, whereas the best 5variable model used all the variables except Ca. As expected, R^{2}values increased as the number of variables in the model increased; however, there was not much improvement in the fit of the regression model when more than 4 variables were in the model.
The regression equations for all the sets of regression models are presented in Table 8. For AME, the full model (the equation that uses all the variables) was AME = 15,190 – (1.22 × GE, kcal/kg) – (5.02 × CP, g/kg) + (33.2 × P, g/kg) – (2.60 × Ca, g/kg) – (5.88 × fat, g/kg) – (16.8 × ash, g/kg). For AME_{n}, the full model was AME_{n} = 15,071 (1.23 × GE, kcal/kg) – (5.02 × CP, g/kg) + (33.8 × P, g/kg) – (2.69 × Ca, g/kg) – (5.56 × fat, g/kg) – (16.9 × ash, g/kg). For AME prediction equation, AIC decreased as the number of variables in the model increased from 1 to 2 and then increased again when the number of variables in the model was >4. For AME_{n}, AIC only decreased when the number of variables in the model increased to 3, and then it increased again when more that 4 variables were in the model. Therefore, on the basis of the values for AIC, the best models were those containing 4 variables, namely GE, CP, P, and ash. The 4variable prediction equations for AME, therefore, was AME = 13,587 – (1.25 × GE, kcal/kg) – (3.51 × CP, g/kg) + (30.4 × P, g/kg) – (16.4 × ash, g/kg), and for AME_{n}, the equation was AME_{n} = 13,547 – (1.25 × GE, kcal/kg) – (3.59 × CP, g/kg) + (31.0 × P, g/kg) – (16.5 × ash, g/kg).
DISCUSSION
The objective of the current pig experiment was to determine the energy value of 21 MBM samples that differ in their chemical composition using digestibility assay and generate regression equations for predicting energy value of MBM. A previous study in our laboratory designed to estimate AME and AME_{n} of MBM samples utilized 12 MBM samples (Adedokun and Adeola, 2005). The use of more samples (21 MBM samples) in the current study was intended to allow for the development of a more robust prediction equation. Also, by using more MBM samples in the current study than the number of samples used in Adedokun and Adeola (2005), the correlation between the proximate fractions and the energy value of the MBM samples would be more reliable. Although nothing appeared to be unique about MBM sample 4, regression model diagnostics using DFBETA criterion revealed that data from this sample were influential points, and therefore, the AME and AME_{n} data from this sample were removed from further regression and correlation analyses. Removal of the data from this MBM sample increased R^{2} from 0.27 to 0.42.
The proximate composition of the MBM samples used in the current study was similar to what had been observed by others (Shi and Noblet, 1993; Vieites et al., 2000). In a previous work in our laboratory, Adedokun and Adeola (2005) reported average fat and P content that are similar to contents reported in the current study. Johnson and Parsons (1997) observed that highash MBM used in their study had decreased GE and that lowash MBM had greater GE, similar to the observation made with the samples used in the current study. In the current study, ash content explained about 95% of the variation in GE content of the MBM samples. High ash content could be because of high bone content. A greater proportion of bones may yield increased CP content in chemical analysis. However, this protein is primarily collagen, which is of little nutritional value to nonruminant animals (Eastoe and Long, 1960; Johnson and Parsons, 1997). This is one of the reasons why addition of an increased quantity of MBM (Peo and Hudman, 1962) or meat meal depressed growth performance in swine (Cromwell et al., 1991).
Batterham et al. (1980) reported that DE of 14 MBM for pigs ranged from 2,247 to 3,322 kcal/kg and Karakas et al. (2001) reported AME_{n} of porcine or bovine MBM in the range of 2,511 to 3,115 kcal/kg for broilers, Waring (1969) reported ME of MBM of 1,988 kcal/kg for colostomized roosters. Adedokun and Adeola (2005) reported AME ranging from 1,569 to 3,308 kcal/kg and AME_{n} ranging from 1,474 to 3,361 kcal/kg for swine.
It can be expected that CP utilization will affect energy utilization of any highprotein feedstuff including MBM. Interestingly, the correlation between GE and DE with the proximate fractions of the MBM was greater (exception was correlation between CP and DE) than the correlation between the same fraction with AME and AME_{n}. In addition, the greatest correlation was between GE and the proximate fractions. What this observation suggests is that it is probably less useful to relate energy utilizability (AME or AME_{n}) with proximate fractions than it is to relate proximate fractions with total content of energy (GE). The main contributors to GE in MBM are CP and fat; each gram of CP and fat contains 5.64 and 9.13 kcal, respectively (Larbier and Leclercq, 1992). Although the GE of fat is greater than that of CP, the CP content of MBM is much greater than its fat content. In the current study, the average proportions of CP and fat were 589 and 113 g/kg, respectively. On average, therefore, the proportional contribution of CP and fat to total GE in each kilogram of the MBM used in the current study was 3.2:1. This was likely why the correlation between CP and GE was greater than the correlation between fat and GE. Because of the greater contribution of CP to GE compared with fat, it is likely that any factor that affects CP utilization will have greater impact on energy utilization than factors affecting fat utilization. Interestingly, the current data showed that as the ratio of CP:fat in MBM increased, energy utilization responses decreased and the correlation was especially stronger for postabsorptive energy utilization measures (i.e., AME and AME_{n)}.
In view of the relationship between the data on energy utilization and fat and CP content of MBM, it is of interest to investigate the factors that may affect the utilization of these proximate fractions and their possible effects on energy utilization. These factors may include AA balance, proportion of saturated:unsaturated fatty acids (Rustan et al., 1993), or ash content (Johnson and Parsons, 1997) especially because ash may affect digesta passage rate or may potentiate interactions between minerals and other nutrients (Atteh and Leeson, 1984). Karakas et al. (2001) used MBM from swine or cattle but with different ash contents in a broiler study and reported that species of origin had no effect on AME_{n} of MBM, but rather, MBM samples with greater ash content (ash content ranged from 21 to 43%) at greater inclusion (20%) depressed AME_{n}. Shirley and Parsons (2001) noted that as the content of ash in MBM increased, there was also a decrease in the level of all essential AA (except Arg) as well as a reduction in protein efficiency ratio of the feedstuff whereas there was no effect on CP digestibility. A decrease in CP utilization coupled with high CP intake will produce increased N load on the animal with consequent increase in expenditure of energy for N excretion and hence a reduction in the amount of available energy to the animal.
Because the N status of an animal receiving a feedstuff will influence the AME of that feedstuff, AME is usually corrected for N retention (AME_{n}). An animal in positive N balance will excrete less N and retain more of its dietary energy than an animal in negative N balance. Noblet et al. (1994) observed in growing pigs that the energy lost in DE to ME step was concerned mainly with digestible CP and digestible ADF and corresponded to the amount of urinary digestible CP energy losses. Urinary N loss is the main route for disposing of excess N arising from catabolism of AA. This N loss, if considerable, represents a significant energy cost to the body.
For example, Diggs et al. (1965) observed that ME was about 82% of DE for pigs receiving highCP feedstuffs compared with the value of about 95% for the same measure in pigs receiving cereal grains. The use of a NE system will likely be more advantageous than ME in describing the energy value of highprotein feedstuffs. We have shown previously (Olukosi et al., 2008) that NE explained more of the variation in performance of broilers than ME. In addition, Just (1982a) noted that NE is the best measure of energy utilization in pigs because it takes into consideration the heat losses associated with digestion and nutrient metabolism. Pirgozliev and Rose (1999) similarly observed in poultry that the efficiency of protein utilization might be the single most important variable to consider in adjusting ME to describe their NE values.
Taverner et al. (1983) pointed out that lysine digestibility of the MBM used in their study with pigs was approximately 50%. In a study with broilers (Karakas et al., 2001), ileal N and total AA digestibility were 55.8 and 56.2%, respectively, and CP digestibility decreased with inclusion level of the MBM. In the same study, the authors reported high correlation (r = 0.9) between AME_{n} and CP content. In the current study, the average digestibility of MBM CP was 83% (data on N utilization are not presented), and N excretion was 18.1 g/d and was not different among the MBM. However, there was a negative correlation between N output and DE, AME, and AME_{n}, and r was more negative (−0.37) for AME and AME_{n} than for DE (−0.14). Although the quantity of N output was also related to CP content (r = 0.48), a comparison of the relationship between N output and energy utilization suggests that the impact of N utilization on energy availability is very important. In fact, except for fat content, the correlation between N output and AME or AME_{n} was greater than the correlation for all the proximate fractions. In addition, the improvement in correlation (from 0.14 to 0.37) between N output and AME or AME_{n} compared with N output and DE further supports the suggestion that CP utilization had great influence on postabsorptive energy utilization. Martosiswoyo and Jensen (1988) similarly reported decreased AME_{n} in broilers receiving greater inclusion level of MBM in their diet, and this was attributed to increased amount of excreted uric acid and N.
The proximate fractions generally explained <50% of the variations in DE, AME, and AME_{n}. However, there were greater correlations between the proximate fractions and GE. In the current experiment, there was a negative correlation between energy and CP and a positive correlation between energy and fat. Adedokun and Adeola (2005) similarly reported negative correlations between CP with AME and AME_{n} of the MBM used in their study. However, the correlation coefficient of fat with AME or AME_{n} in Adedokun and Adeola (2005) was greater than observed in the current study, whereas the opposite was true for CP content of the MBM. The difference between the results obtained may be related to the MBM samples used. The MBM samples used in the current study had greater GE and CP and less Ca and ash than the ones used in Adedokun and Adeola (2005). However, results of the current study indicate that the proximate fractions did not explain much of the variation in energy utilization.
We are not aware of any study reporting the correlation of ratio of energy yielding fractions with one another or with ash on AME or AME_{n} of MBM. However, other studies have shown that the ratio of CP or proportions of calories contributions from energyyielding nutrients can influence energy value of feedstuffs. For example, Hartsook et al. (1973) observed that proportional calories contribution from carbohydrate and fat decreased DE and ME of rat diets at a reduced CP level but had the opposite effect at an increased CP level. Just (1982b) observed that increasing the level of crude fiber in swine diet depressed ME by shifting digestion to the more distal portion of the digestive tract. However, Just (1982a) reported that increasing concentration of fat in diet for growing pigs increased the ME values of the diets because of greater efficiency of absorption of fatty acids as concentration of fats in the diets increased. Consequently, we were interested in examining how the ratios of energyyielding nutrients in MBM may relate to AME and AME_{n}. However, in the current study, correlations of the ratio of proximate fractions did not explain more of the variations than were explained by the individual proximate fractions.
Because of the wide variation in the composition, species of origin, and processing techniques of MBM, chemical composition, energy content, and energy utilization will vary widely. It is not possible to determine the energy value of all possible types of MBM available, but it is hoped that an examination of MBM obtained from different species and having different chemical composition may yield a robust prediction equation that may be applicable in many situations. In the current study, to establish the best prediction equation, the full and reduced models were tested to establish which of the predictors can be eliminated from the prediction equation. The Pvalues for the partial correlations did not reach significance (at 5% probability), indicating that there was no substantial gain in precision by reducing the number of predictors used in the regression equation. Consequently, the decision to eliminate some of the predictors was based on the result obtained from the modelbuilding process using multiple layers of elimination criteria (R^{2}, AIC, and Cp) to identify the best model candidate.
Batterham et al. (1980) noted that the best relationship between DE of MBM and the proximate compositions required inclusion of GE, fat, Ca, and P in the model. Adedokun and Adeola (2005) described a model that incorporated CP and ash for predicting AME and AME_{n} of MBM for pigs. As explained earlier, regression coefficient will increase as more variables are added to the model and so it is not a very useful criterion for choosing the optimum model. However, AIC penalizes for complexity of model and therefore ensures the use of the minimum number of variables. There is no perfect model, and because the objective was to select the best model from a pool of possible model subsets, a combination of selection criteria was used in the current study.
In the current experiment, the best model for predicting AME and AME_{n} of MBM for swine used 4 variables, namely GE, P, CP, and ash. An examination of the information provided by R^{2}, Cp, and AIC helped inform the choice of this model as previously explained in the Results section. Adedokun and Adeola (2005) noted that the best reduced model incorporated only CP and ash in MBM for predicting AME_{n} for swine, but the choice of the model in that study was based on R^{2} and SD values. Although the MBM used in the 2 studies were different, the chemical compositions were similar (the differences are noted earlier). However, in the 2 studies, ash and CP were included in the reduced model. It is likely that the use of more selection criteria in the current study allowed the choice of a model that is more precise and at the same time eliminated redundant variables that did not add significantly to the precision of the prediction equation.
In conclusion, the result of this study showed that MBM is, in addition to being a source of CP and minerals, a good energy source with an average AME value of 3,070 kcal/kg. In addition, the current study highlighted the potential of using proximate compositions for predicting the AME and AME_{n} of MBM for pigs. The study also established that in addition to the proximate compositions, factors that influence the utilization of these fractions, especially fat and CP, may have substantial impact on energy utilization of MBM. We believe that although other factors extrinsic to MBM may have influence on its energy utilization, the use of proximate fractions as well as the characteristics of these fractions should be sufficient for predicting the energy value of MBM for swine.
Ingredient  
Corn  717.0  661.0 
Soybean meal  245.0  226.0 
Dicalcium phosphate  16.0  0.0 
Limestone  9.0  0.0 
Salt  3.0  3.0 
Vitamin premix^{1}  3.0  3.0 
Selenium premix^{2}  0.5  0.5 
Trace mineral premix^{3}  1.5  1.5 
Lysine·HCl  2.5  2.5 

1.0  1.0 
Threonine  0.7  0.7 
Tryptophan  0.3  0.3 
Antioxidant^{4}  0.5  0.5 
Meat and bone meal  0.0  100.0 
Total  1,000  1,000 
Calculated nutrient and energy^{5}  
CP, g/kg  175.9  162.2 
DE, kcal/kg  3,430  3,163 
ME, kcal/kg  3,280  3,025 
Ca, g/kg  7.2  11.0 
P, g/kg  6.7  8.4 
1  960  4,269  548  47.2  94.8  104.9  284.7 
2  939  4,657  641  37.9  70.9  100.7  239.7 
3  952  4,167  571  51.2  102.9  105.3  285.3 
4  946  4,605  601  37.2  80.2  119.4  247.7 
5  939  4,270  612  46.8  93.4  92.1  280.4 
6  956  3,895  491  56.0  116.2  120.4  338.2 
7  965  3,968  514  51.8  107.1  120.9  328.2 
8  961  4,722  631  36.8  66.8  101.3  217.1 
9  965  4,769  629  32.4  66.2  106.8  221.6 
10  964  4,761  632  35.8  64.5  102.5  226.6 
11  966  4,734  554  37.5  76.9  108.8  241.3 
12  963  4,720  626  38.5  72.5  112.3  224.7 
13  967  4,789  621  38.8  72.6  110.9  242.0 
14  967  4,702  619  38.0  77.8  115.1  231.1 
15  942  5,077  570  23.1  38.0  135.5  142.0 
16  944  5,106  569  23.9  40.5  135.1  148.0 
17  945  5,193  573  22.5  37.0  141.2  147.3 
18  971  4,640  570  44.2  97.0  110.8  263.9 
19  953  4,627  626  41.7  79.3  105.3  227.1 
20  969  4,247  567  29.5  86.6  106.0  263.9 
21  967  4,697  594  35.4  69.8  116.0  210.7 
Essential AA  
Arginine  35.8  40.5  39.0  40.0  40.0  33.5  34.7  37.6  37.9  37.7  37.4  39.1  38.5  38.0  34.4  34.0  34.5  37.7  41.6  36.0  38.0 
Histidine  10.6  16.2  10.0  13.5  12.1  8.2  8.5  14.8  14.9  15.0  13.9  14.4  14.4  14.0  13.7  13.0  13.7  11.3  11.1  12.9  14.2 
Isoleucine  16.3  22.6  15.7  22.4  18.5  13.2  13.6  21.6  21.9  21.0  20.8  20.9  20.7  20.6  21.2  20.2  21.3  17.4  20.1  19.6  20.8 
Leucine  33.4  44.0  33.3  42.4  37.2  28.1  29.4  41.2  41.0  40.1  39.3  39.8  39.4  38.6  39.6  37.9  39.4  35.2  39.8  37.7  40.0 
Lysine  30.8  40.6  29.2  39.1  33.3  24.7  24.2  38.1  38.5  38.7  37.0  38.2  38.0  37.4  31.8  30.7  32.0  30.7  29.9  33.3  35.8 
Methionine  7.3  10.9  7.7  11.3  8.4  6.4  6.1  10.0  9.9  9.8  9.8  9.8  9.6  9.6  8.9  8.8  9.1  8.4  8.2  9.7  9.8 
Phenylalanine  18.4  23.6  18.1  22.8  20.6  15.7  16.3  22.3  22.0  21.8  21.2  21.4  21.3  20.9  21.8  20.9  21.6  19.5  22.2  20.8  21.7 
Threonine  16.9  20.8  17.4  21.9  19.0  14.4  14.3  21.9  21.1  21.1  20.5  20.4  20.2  19.2  19.9  19.0  19.3  17.3  20.0  18.7  20.1 
Tryptophan  4.2  5.0  3.7  5.3  4.5  3.3  3.0  4.8  4.9  4.9  4.1  4.6  4.3  4.4  3.8  3.8  4.6  3.6  3.5  4.6  4.9 
Valine  24.1  31.3  23.6  29.5  27.4  19.9  21.1  28.0  28.4  26.9  27.2  27.4  27.2  27.1  27.6  26.6  27.6  24.9  30.3  26.9  28.0 
Nonessential AA  
Alanine  39.3  45.0  39.0  41.3  42.8  35.7  38.4  41.7  41.6  42.1  42.1  43.0  42.6  42.1  36.3  35.8  36.2  41.6  43.3  37.8  40.7 
Aspartic acid  39.9  49.6  40.1  49.2  43.8  34.4  35.3  49.5  48.3  47.9  47.8  47.1  46.6  46.0  44.6  43.4  44.6  41.8  44.3  43.8  46.1 
Cysteine  4.7  4.9  6.8  6.1  6.6  4.8  3.8  4.6  5.6  5.2  4.5  5.1  5.1  4.8  6.4  6.5  6.7  4.3  8.4  5.9  5.4 
Glutamic acid  64.5  80.4  68.2  76.8  71.6  57.1  59.3  76.9  76.9  79.3  74.8  75.2  74.7  73.8  76.2  72.8  74.2  69.0  72.3  69.7  74.4 
Glycine  67.2  64.0  71.0  60.7  72.9  66.8  71.8  60.9  59.5  61.8  64.3  65.4  64.4  64.7  49.5  51.0  49.9  69.8  73.8  56.0  59.7 
Hydroxylysine  3.0  2.6  2.7  2.5  3.1  2.8  3.1  2.4  2.6  2.5  2.7  2.8  2.7  2.8  1.9  2.1  2.1  3.2  3.1  2.4  2.8 
Hydroxyproline  26.3  21.0  27.5  20.5  27.4  27.3  29.3  21.5  21.1  23.3  23.3  24.4  23.8  23.4  14.5  15.7  15.0  26.5  25.8  18.6  20.1 
Lanthionine  0.1  0.1  3.3  0.1  2.3  0.7  0.1  0.0  0.1  0.2  0.1  0.1  0.1  0.1  0.1  0.2  0.1  2.2  3.6  0.1  0.1 
Ornithine  0.7  0.9  0.5  0.4  0.7  0.4  0.4  0.9  0.8  0.6  1.0  0.8  0.8  0.8  0.6  0.8  1.0  0.8  1.0  0.6  0.7 
Proline  42.3  42.5  45.6  40.3  47.1  39.2  42.4  40.9  39.6  44.5  46.4  42.1  41.3  41.4  32.7  35.6  34.9  41.2  49.4  37.4  37.3 
Serine  18.4  19.3  20.9  21.8  22.2  17.1  16.2  21.3  20.6  21.2  20.1  19.9  19.8  18.3  19.6  18.6  18.4  17.7  25.4  18.2  19.3 
Taurine  0.8  1.4  0.7  0.6  0.7  0.6  0.5  0.9  1.0  0.8  1.0  1.2  1.1  1.0  1.3  1.3  1.4  0.9  0.8  1.2  1.0 
Tyrosine  12.9  16.2  12.6  16.5  12.9  10.1  10.3  15.2  15.2  15.0  14.4  14.7  14.4  14.0  14.9  14.2  14.6  13.5  14.8  13.7  14.6 
Total  517.7  602.8  536.8  585.6  575.1  464.4  481.9  577.5  573.1  580.9  573.5  577.4  570.8  562.6  521.2  512.7  521.7  538.6  592.9  525.3  555.3 
1  3,518  3,384  3,283 
2  3,267  2,901  2,786 
3  3,996  3,890  3,794 
4  4,120  3,842  3,733 
5  3,185  2,840  2,729 
6  2,669  2,611  2,512 
7  3,241  3,101  3,001 
8  3,387  3,001  2,889 
9  2,792  2,613  2,510 
10  3,125  2,714  2,600 
11  3,576  3,210  3,102 
12  3,534  3,267  3,170 
13  2,704  2,320  2,212 
14  3,367  2,988  2,881 
15  3,512  3,160  3,053 
16  3,863  3,442  3,341 
17  4,176  3,872  3,767 
18  3,130  2,627  2,517 
19  3,385  2,724  2,609 
20  3,353  2,860  2,745 
21  3,552  3,101  2,994 
SEM  207.7  219.7  222.7 
AME  —  0.99  0.92  0.15  −0.25  −0.20  −0.25  0.41  −0.29 
AME_{n}  0.93  0.15  −0.26  −0.19  −0.24  0.42  −0.29  
DE  0.34  −0.07  −0.38  −0.40  0.40  −0.48  
GE  0.50  −0.86  −0.94  0.44  −0.95  
CP  −0.34  −0.42  −0.44  −0.42  
Phosphorus  0.93  −0.49  0.91  
Calcium  −0.50  0.98  
Fat  −0.52  
Ash  — 
AME  0.36  0.44  0.33  −0.39  0.39  −0.33 
AME_{n}  0.36  0.44  0.32  −0.39  0.39  −0.33 
DE  0.49  0.53  0.48  −0.29  0.43  −0.17 
GE  0.90  0.78  0.91  −0.05  0.58  0.29 
Apparent ME, kcal/kg  
1  0.17  2.54  Fat 
1  0.09  4.47  Ash 
2  0.26  2.59  CP, ash 
2  0.26  2.65  GE, ash 
3  0.34  2.80  GE, CP, ash 
3  0.32  2.23  GE, P, ash 
4  0.41  3.14  GE, P, CP, ash 
4  0.37  4.14  GE, P, fat, ash 
5  0.42  5.02  GE, P, CP, fat, ash 
5  0.41  5.13  GE, CP, P, Ca, ash 
Nitrogencorrected apparent ME, kcal/kg  
1  0.17  2.51  Fat 
1  0.08  4.55  Ash 
2  0.26  2.66  CP, ash 
2  0.26  2.68  GE, ash 
3  0.34  2.83  GE, CP, ash 
3  0.32  3.28  GE, P, ash 
4  0.41  3.12  GE, P, CP, ash 
4  0.37  4.12  GE, P, fat, ash 
5  0.42  5.02  GE, P, CP, fat, ash 
5  0.42  5.12  GE, CP, P, Ca, ash 
Apparent ME, kcal/kg of DM  
1  240  0.17  1,530  13.3  
2  239  0.26  13,037  −1.54  −12.2  
3  239  0.34  12,727  −1.14  −3.24  −10.7  
4  239  0.41  13,587  −1.25  −3.51  30.4  −16.4  
5  241  0.42  14,994  −1.20  −4.90  31.6  −5.49  −16.1  
6  242  0.42  15,190  −1.22  −5.02  33.2  −2.60  −5.88  −16.8 
Apparent nitrogencorrected ME, kcal/kg of DM  
1  240  0.17  1,401  13.5  
2  240  0.26  6,451  −4.49  −3.72  
3  239  0.34  12,670  −1.14  −3.31  −10.7  
4  239  0.41  13,547  −1.25  −3.59  31.0  −16.5  
5  240  0.42  14,870  −1.21  −4.90  32.2  −5.16  −17.5  
6  242  0.42  15,071  −1.23  −5.02  33.8  −2.69  −5.56  −16.9 