Purdue University 2000 Swine Day Report

The Ability of Measures of Carcass Composition to Predict Empty Body Protein and Lipid Composition

A.P. Schinckel, J.R. Wagner, J.C. Forrest, and M.E. Einstein
Department of Animal Sciences and Elanco Animal Health, Indianapolis

Introduction

Pork carcass and empty body composition research has been conducted to evaluate effects of experimental treatment, to model pig growth, and to evaluate pork production systems. Scientists with an interest in modeling pig growth and predicting nutrient requirements require accurate esimates of empty body chemical composition. Empty body protein and lipid mass are expensive and difficult to obtain. Actual or predicted measures of fat-free lean are more easily obtained.

The objectives of this study were to evaluate the reliability of predictive measures of carcass composition and to develop a further understanding of the interrelationships among various pork carcass and empty body composition endpoints.

Materials and Methods

Data from 203 pigs, representing seven genotypes and two sexes (barrows and gilts), were used to evaluate relationships among different measures of carcass composition. Details of the experimental design and data are presented in Schinckel et al. (2000). Pigs were slaughtered at four target weights: 220, 251, 282, or 334 lbs. The data were analyzed as two separate data sets: a light weight data set (target weights of 220, 251, and 282 lbs) and a heavy weight data set (target weights of 251, 282, and 334 lbs).

The ability to accurately predict empty body protein (MTPRO) and lipid mass (MTLIP) from the carcass measures was evaluated by regression analyses. MTPRO data were fit to regression equations including carcass weight (CW) and either the actual or predicted values of fat-free lean mass (FFLM), lipid-free soft tissue (LFSTIS), or dissected lean in the four lean cuts (DL). Empty body lipid mass (MTFAT) data were fit to regression equations including CW and either the actual or predicted values of total carcass fat (TOFAT) and total soft tissue lipid (TLIPID). The predicted values of the alternative measure of carcass composition were predicted from ribbed carcass measurements.

Accuracy of each prediction equation was evaluated by R2, which is the multiple coefficient of determination, and the residual standard deviation (RSD). Least squares means of the residual values for the genetic population, sex, and target weight subclasses were evaluated as estimates of subpopulation biases (Gu et al., 1992). The correlation coefficients (CR) between the predicted and observed genotype-sex means were used as measures of genotype bias. The proportion of variation among genotypes accounted for by each equation was determined by the variance ratio (VR), which is the variance of predicted genotype-sex means divided by the variance of observed means.

Results

Acronyms and definitions of the variables are presented in Table 1. Table 1 also contains overall and sex means for the light and heavy weight data sets. Table 2 contains means for empty body protein and empty body lipid mass, for the two sexes, four weight groups, and seven genetic populations.

Prediction equations for MTPRO are presented in Table 3 and the residual value statistics are presented in Table 4. Equation 1 includes the standard ribbed carcass measurements (CW, 10th rib fat depth [FD10R] and carcass loin eye area [LEA]). The other equations include CW (if significant) and either the actual or predicted value of LFSTIS or FFLM. The predicted values were determined from equations including the ribbed carcass measurements.

Overall, there was no substantial advantage in terms of R2 or magnitude of biases to favor the use of the actual versus the predicted LFSTIS or FFLM in the prediction of MTPRO. Also, there was no advantage of using one measure of carcass "lean" mass over another in the prediction of MTPRO. Carcass weight was significant for each equation (P<.01), indicating that MTPRO cannot be predicted as a simple linear function of any of the actual or predicted carcass composition measurements. Not including the CW term resulted in significant (P<.01) weight group biases.

LEA was not significant (P=.92) in equation 1 for the light weight data and had marginal impact (P=.17) in the heavy pig data. The MTPRO of genetic population 4 (G4) was overestimated by each prediction equation. This is likely caused by G4 having a different ratio of MTPRO to the carcass composition measures than the other genetic populations. G4 had less visceral protein than the overall mean (3.77 vs. 4.03 lb) and a slightly higher dressing percentage (75.8 vs. 74.8%). For all other genetic populations, the mean absolute value of bias is 1.0 to 1.7% of the mean value. Sex and weight group biases were not significant in the prediction of MTPRO.

Prediction equations for MTLIP are presented in Table 5 and the summary of residual value statistics are presented in Table 6. Overall, the R2 are higher (.88 to .92) for the prediction equations for MTLIP than MTPRO (R2 = .74 to .82). Overall, the use of actual TOFAT or STLIP data in the prediction of MTLIP resulted in R2 values of .82 to .93. The inclusion of predicted TOFAT and STLIP of standard carcass measurements resulted in R2 values of .88. Carcass weight was significant for each equation (P<.001), indicating that MTLIP cannot be predicted as a simple linear function of the actual or predicted values of TOFAT or STLIP. The regression coefficients for the predicted or actual values of TOFAT and STLIP are smaller for the heavy weight data than the light weight data.

There were no significant genetic population (P>.10) or weight group (P>.57) biases in the prediction of MTLIP. The CR values (.97 to .98) indicate that the prediction equations rank the genetic populations correctly. The VR values ranged from 85 to 93%. However, the MTLIP of gilts was over predicted by .73 lb in the light weight and .95 lb in the heavy weight data. With an 8.34 and 10.23 lb difference in MTLIP between the barrows and gilts in the light and heavy weight data (Table 1), the prediction equations account for 80% of the true difference in MTLIP between the barrows and gilts.

Discussion

The mass and growth of MTPRO and MTLIP are primary inputs needed to model nutrient (energy, protein and macromineral) requirements. MTPRO and MTLIP are expensive to obtain on a regular basis; thus, they must be predicted from less costly measurement methods in order to estimate farm-specific nutritional requirements.

It was expected that the measures of carcass composition based on chemical analysis (STLIP and LFSTIS) would result in the most accurate predictions of MTPRO and MTLIP. Standard ribbed carcass measurements (CW, LEA, FD10R) and either the actual or predicted values of LFSTIS or FFLM resulted in MTPRO prediction equations with similar accuracy in terms of RSD, VR, CR and magnitude of biases. Equations including STLIP were slightly more accurate (.01 higher R2, .1 to .2 lower RSD) in predicting MTLIP than equations including TOFAT. Equations including the actual STLIP or TOFAT values were more accurate (.04 to .05 higher R2, .5 to .7 lb lower RSD) than equations using the predicted values of TOFAT, STLIP or ribbed carcass measurements.

The predicted values of the measures of carcass composition were from equations including CW, FD10R and LEA. These equations were the most accurate in terms of RSD, VR and CR. Using the predicted values from other prediction equations would have resulted in less accurate, more biased prediction of the measures of empty body composition.

Implications

Empty body protein and lipid mass are used to predict daily nutrient requirements. Predicted or actual values of measures of carcass composition or standard carcass measurements can be used to predict empty body protein and lipid in market weight pigs.

References

Gu, Y., A.P. Schinckel, T.G. Martin, J.C. Forrest, C.H. Kuei, and L.E. Watkins. 1992. Genotype and treatment biases in estimation of carcass lean of swine. J. Anim. Sci. 70:1708-1718.

Schinckel, A.P., J.R. Wagner, J.C. Forrest, and M.E. Einstein. 2000. Evaluation of alternative measures of pork carcass composition. Purdue University Swine Day Report. p. 105.

 

Table 1. Overall barrow and gilt means for the light and heavy pig data sets.

 

 

220, 251, and 284 lb
weight groups

251, 284, and 334 lb
weight groups


Acronym

Definition of variable and
unit of measurement

Overall mean


Barrows


Gilts


SD

Overall mean


Barrows


Gilts


SD

LW

Live weight, lb

247.70

247.60

247.80

26.5

284.20

284.60

283.80

34.4

CW

Warm carcass weight, lb

185.01

187.20

185.70

22.5

214.62

214.83

214.4

27.8

FFLM

Fat-free lean mass, lb

80.93

77.00

84.75

12.1

90.10

86.42

95.46

13.4

TOFAT

Total carcass fat tissue mass, lb

65.05

73.34

64.97

15.2

84.81

90.79

78.93

20.5

LFSTIS

Lipid-free soft tissue mass, lb

98.34

94.82

101.70

12.3

111.55

107.87

115.17

15.0

STLIP

Carcass soft tissue lipid mass, lb

51.68

55.53

40.02

13.4

64.24

69.36

59.22

16.7

MTPRO

Empty body protein, lb

31.92

30.62

33.25

4.0

35.67

34.13

37.26

5.1

MTLIP

Empty body lipid, lb

68.94

73.24

64.90

15.6

85.16

90.32

80.09

19.1

 

Table 2. Least squares means for empty body protein mass and empty body lipid mass in the dissected fat tissuea.

 

Empty body
protein, lb

Empty body
lipid, lb

Sex

 

 

Barrows

32.41

82.23

Gilts

35.27

73.63

SE

.25

.88

Significance

.0001

.0001

 

Weight group

 

 

220

28.22

54.90

251

31.75

67.46

282

35.71

84.44

334

40.56

105.2

SE

.40

1.26

Significance

.0001

.0001

 

Genetic population

 

 

1

32.63

83.11

2

35.94

70.99

3

34.61

77.16

4

33.73

79.81

5

33.29

80.25

6

33.51

80.69

7

34.39

73.85

SE

.49

1.59

Significance

.001

.0001

aPooled dissected lean or fat tissue from the four lean cuts.

 

Table 3. Equations and regression analysis for predicting empty body protein mass (lb) using various carcass measurements and dissection data.

 

 

220, 251, and 282 lb weight groupsb (N=152)

251, 292, and 334 lb weight groupsb (N=152)

Eq.

Variablea

R2

RSD, lb

b0

bi

Signifc

R2

RSD, lb

b0

bi

Signifc

1

CW

.75

.93

8.18

.167

.001

.79

1.08

4.06

.170

.001

 

LEA

 

 

 

(NS)

.92

 

 

 

.307

.17

 

FD10R

 

 

 

-5.88

.001

 

 

 

-5.28

.001

 

2

CW

.75

.93

2.05

.0324

.01

.81

1.01

-.86

.0461

.001

 

LFSTIS

 

 

 

.242

.001

 

 

 

.238

.001

 

3

CW

.74

.95

1.63

.0191

.17

.79

1.08

-1.08

.0415

.001

 

PLFSTIS

 

 

 

.272

.001

 

 

 

.248

.001

 

4

CW

.77

.89

3.77

.0578

.001

.81

1.00

1.87

.0851

.001

 

FFLM

 

 

 

.224

.001

 

 

 

.212

.001

 

5

CW

.74

.95

3.61

.0475

.001

.79

1.08

-2.31

.0808

.001

 

PFFLM

 

 

 

.240

.001

 

 

 

.227

.001

a CW = warm carcass weight (lb), LEA = 10th rib loin eye area (in2), FD10R = off-midline 10th rib fat depth (in), LFSTIS = lipid-free soft tissue, PLFSTIS = predicted lipid-free soft tissue mass (lb), and PFFLM = predicted fat-free lean muscle mass (lb). PFFLM and PLFSTIS were predicted using equations which included CW, LEA, and FD10R.
b b0 = intercept, bi = partial regression coefficient of the ith independent variable, R2 = coefficient of determination, RSD = residual standard deviation.
c Signif = significance.

 

Table 4. Significance of residual mean value (biases) for genotype populations and sex subpopulations in the prediction of empty body protein mass.a

 

 

Light weight data sets
(220, 251, and 282 lbs)

Heavy weight data sets
(251, 282, and 335 lbs)



Eq.



Variable

Genetic
Population
Signif.


Sex
Signif.



CRb



VRb

Genetic
Population
Signif.


Sex
Signif.



CRb



VRb

1

CW, LEA, FD10R

.11

.54

.86

.79

.41

.34

.90

.78

2

CW, LFSTIS

.06

.40

.83

.76

.50

.19

.91

.84

3

CW, PLFSTIS

.05

.68

.82

.79

.35

.37

.89

.79

4

CW, FFLM

.15

.62

.88

.78

.29

.38

.93

.90

5

CW, PFFLM

.05

.68

.83

.79

.34

.38

.89

.79

a CW = warm carcass weight (lb), LEA = 10th rib loin eye area (in2), FD10R = off-midline 10th rib fat depth (in), FFLM = fat-free lean mass (lb), PFFLM = predicted fat-free lean muscle mass (lb), LFSTIS = lipid-free soft tissue mass (lb), PLFSTIS = predicted lipid-free soft tissue mass (lb), DL = dissected lean in the four lean cuts (lb), and PDL = predicted dissected lean mass (lb).
b CR and VR are the correlations and variance ratios, respectively, between the predicted and observed genetic population-sex means.

 

Table 5. Equations and regression analysis for predicting empty body lipid mass (lb) using various carcass measurements and dissection data.

 

 

220, 251, and 282 lb weight groupsb (N=152)

251, 292, and 334 lb weight groupsb (N=152)

Eq.

Variablea

R2

RSD, lb

b0

bi

Signifc

R2

RSD, lb

b0

bi

Signifc

1

CW

.88

2.5

-32.8

.465

.001

.88

3.0

-34.2

.489

.001

 

LEA

 

 

 

-2.06

.008

 

 

 

-2.36

.005

 

FD10R

 

 

 

21.95

.001

 

 

 

20.55

.001

 

2

CW

.92

2.0

-13.7

.134

.001

.92

2.5

-7.1

.133

.001

 

TOFAT

 

 

 

.834

.001

 

 

 

.749

.001

 

3

CW

.88

2.5

-12.6

.108

.001

.88

3.0

-4.6

.0977

.005

 

PTOFAT

 

 

 

.890

.001

 

 

 

.811

.001

 

4

CW

.93

1.9

-5.3

.137

.001

.93

2.3

-8.8

.177

.001

 

STLIP

 

 

 

.945

.001

 

 

 

.872

.001

 

5

CW

.88

2.5

-4.6

.121

.001

.88

3.0

-8.8

.174

.001

 

PSTLIP

 

 

 

.986

.001

 

 

 

.877

.001

a CW = warm carcass weight (lb), LEA = 10th rib loin eye area (in2), FD10R = off-midline 10th rib fat depth (in), TOFAT = total carcass fat mass (lb), PTOFAT = predicted total carcass fat mass (lb), STLIP = soft tissue lipid mass (lb), and PSTLIP = predicted soft tissue lipid mass (lb). PTOFAT and PSTLIP are predicted using equations which included CW, LEA, and FD10R.
b b0 = intercept, bi = partial regression coefficient of the ith independent variable, R2 = coefficient of determination, RSD = residual standard deviation.
c Signif = significance.

 

Table 6. Significance of residual mean value (biases) for genotype populations and sex subpopulations in the prediction of empty body lipid mass.a

 

 

Light weight data sets
(220, 251, and 282 lbs)

Heavy weight data sets
(251, 282, and 335 lbs)



Eq.



Variable

Genetic
Population
Signif.


Sex
Signif.



CRb



VRb

Genetic
Population
Signif.


Sex
Signif.



CRb



VRb

1

CW, LEA, FD10R

.46

.08

.97

.85

.87

.06

.97

.80

2

CW, TOFAT

.40

.09

.98

.93

.12

.12

.97

.81

3

CW, PTOFAT

.48

.08

.97

.85

.90

.05

.97

.80

4

CW, STLIP

.50

.09

.98

.92

.90

.12

.98

.89

5

CW, PSTLIP

.46

.08

.97

.85

.90

.05

.97

.80

a CW = warm carcass weight (lb), LEA = 10th rib loin eye area (in2), FD10R = off-midline 10th rib fat depth, TOFAT = total carcass fat tissue mass (lb), PTOFAT = predicted total carcass fat tissue mass (lb), STLIP = soft tissue lipid (lb), and PSTLIP = predicted soft tissue lipid mass (lb). PTOFAT and PSTLIP were predicted using equations which included CW, LEA, and FD10R.
b CR and VR are the correlations and variance ratios, respectively, between the predicted and observed genetic population-sex means.


Index of 2000 Purdue Swine Day Articles