R .E. Agnew, T. Yan Livestock Production Science 66 2000 197 –215
207
4. Energy value per unit of liveweight change Bath et al., 1964; Reid and Robb, 1971; Tamminga
et al., 1997. Protein mobilisation is stopped at about Energy value per unit of liveweight change for
4 weeks after calving, while fat mobilisation is not lactating dairy cows is fixed in NRC system and each
stopped at 8 weeks Tamminga et al., 1997. This of European systems, but varies with a range be-
may be due to a considerable hypertrophy of both tween 19 and 30 MJ kg of liveweight change. A
gut and liver during early lactation in response to number of recent studies have reported different
increased feed intake Reynolds and Beever, 1995. energy values within the above range Chilliard et
The composition of mobilised tissue fat and protein al., 1991; Gibb et al., 1992; Tamminga et al., 1997.
can thus differ during the first 4 weeks of lactation, The reason for this discrepancy may be attributed to
i.e. the ratio of fat-to-protein would increase. If the the effects of body condition and lactation stage.
water content in the mobilised tissue is assumed to Body condition factors have been recognised to
remain unchanged, the energy value per unit of influence the energy value of liveweight change.
liveweight loss would be higher as lactation pro- These factors include gain or loss of body fat or
gresses. On the other hand, energy balance is not protein, replacement of body fat with water and
always related to liveweight change of dairy cows. changes in gut fill. The lack of precision in these
Beever et al. 1998 reported that high genetic merit factors can result in errors in ration formulation and
cows were still in negative energy balance at 20 the prediction of animal performance, especially in
weeks of lactation, but liveweight of the animals was early and late lactation when liveweight change of
maintained after 5 weeks. This can be partially dairy cattle can be large. It may not be realistic in
explained by the findings of Tamminga et al. 1997 practice to distinguish the effects of gut fill on gain
as changes in the water content in mobilised or loss of liveweight and how much water is
liveweight. The total water loss was proportional to contained in the weight of change. However, a
the mobilised liveweight during the first 4 weeks of number of studies have related the energy value of
lactation. Afterwards the animals retained water in liveweight change to body condition. In a serial
their bodies and liveweight loss was very small, but slaughter study with Holstein–Friesian cows, Gibb
energy loss was still high Tamminga et al., 1997. and Ivings 1993 reported that the body fat and
Energy value per unit of mobilised liveweight thus energy contents of animals were positively related to
increased substantially from 1 to 8 weeks of lactation their liveweight and condition score and body protein
Tamminga et al., 1997. content was positively related to their liveweight.
The fixed energy value for liveweight change in Fat-free mass of Holstein cows remains similar
NRC system and European systems is therefore during the dry period, early and late lactation stages,
incorrect and can result in errors in practice, espe- while water content of fat-free mass is greater during
cially for high genetic merit cows which have a high the dry period and early lactation than late lactation
liveweight loss during early lactation. There is a stages Andrew et al., 1994. These relationships are
need to develop more appropriate measures of reflected in the Cornell net carbohydrate and protein
energy status in dairy cows. If liveweight change system Fox et al., 1992. In this system the energy
continues to be used as an index of tissue energy concentration per condition score is linearly and
change, it must be related to both body condition and positively related to body condition score and
stage of lactation. liveweight of dairy cattle. In Australia a linear
regression equation for dairy cattle has been de- veloped to relate the energy value of liveweight
5. Validation of the UK ME system and other
change to body condition score SCA, 1990. energy systems
Another factor
affecting energy
value per
liveweight change is stage of lactation. During early The data used in the present validation are the
lactation dairy cows can mobilise both fat and experiment mean data n 542 derived from the
protein from their body reserves for milk production, literature, used to develop Eqs. 1 and 2, as
but protein mobilisation decreases faster than fat described previously for references see Appendix
208 R
.E. Agnew, T. Yan Livestock Production Science 66 2000 197 –215
1. The energy systems examined include the UK 1
2
] ME system AFRC, 1990 and 1993, the Australian
MSPE 5
O
A 2 P 5
n ME system SCA, 1990, the American NE system
] ]
2 2
2 2
2
NRC, 1988 and the Dutch NE system Van Es, MSPE 5 A 2P 1 S 1 2 b 1 S 1 2 R
6
P A
1978. AFRC 1993 is a working version of AFRC 1990 which adds a proportionately 0.05 to total
where n is the number of pairs of values of A and P ]
] ME requirement predicted from AFRC 1990. The
compared n 542; A and P are, respectively, the
2 2
Dutch NE system was chosen to represent a range of means of A and P; S
and S are, respectively, the
P A
the NE systems based on the principles proposed by variances of A and P; b and R are, respectively, the
Van Es 1975, e.g. used in France INRA, 1989, slope and correlation coefficient of the linear regres-
the Netherlands Van Es, 1978, Switzerland Bickel sion of A on P. The three components are thus due to
] ]
and Landis, 1978 and Germany Deutschen Land- mean bias A2P, line bias the deviation of the
wirtschafts Gesellschaft, 1982. The American NE slope and random variation of the slope. In order to
system NRC, 1988 does not provide any method to compare the accuracy of prediction between ME and
calculate intake of NE for lactation NEL from its NE systems, the MSPE is expressed as the mean
]] Œ
¯ ME intake, so the observed NEL intake in the
prediction error MPE 5 MSPE A . present validation for the NRC system is based on
The results of the present validation are presented the equation of Moe 1981 NEL50.697 ME2
in Table 3 and Fig. 4a and b. The MPE was much 0.877, MJ kg DM. The predicted NEL requirement
higher with AFRC 1990 than the other three in Van Es 1978 and NRC 1988 is a sum of
systems Van Es, 1978; NRC, 1988; SCA, 1990, predicted NE
and observed E and E . The pre- indicating that AFRC 1990 is the least accurate at
m l
g
dicted ME requirement is a sum of predicted NE predicting total energy requirements of lactating
m
and observed E and E divided by relevant energetic dairy cows. The predicted error of AFRC 1990 was
l g
efficiencies. In both the ME and NE systems E was largely derived from an under-prediction of total
g
] ]
treated as, if E was positive then it was added; if E energy requirement mean bias, A2P, which re-
g g
was negative then it was subtracted after multiplica- sulted in a large proportion of mean bias over total
tion by either 0.80 Van Es, 1978, 0.82 NRC, MSPE. However, when a proportionately 0.05 was
1988, or 0.84 AFRC, 1990; SCA, 1990. added to total ME requirement predicted in AFRC
The model, used to compare those energy systems 1990, as suggested by AFRC 1993 currently
for actual energy ME or NE intake A and adopted in UK, total energy requirements were
predicted energy requirement P, is the mean-square predicted with a similar accuracy to NRC 1988 and
prediction error MSPE as described by Rook et al. SCA 1990. Van Es 1978 had a marginally higher
1990. The MSPE is defined as Eq. 5 and can be MPE than NRC 1988, SCA 1990 and AFRC
regarded as the sum of three components Eq. 6 1993. The higher prediction error for Van Es
Table 3 Prediction precision of different energy feeding systems in calorimetric data of dairy cows published since 1976 n 542
Systems Energy intake MJ day
MSPE MPE
Proportion of MSPE Actual
Pred. Bias
Bias Line
Random ME system
AFRC 1990 181
171 9.9
219 0.082
0.45 0.01
0.54 AFRC 1993
181 179
1.4 129
0.062 0.01
0.07 0.92
SCA 1990 181
182 2
1.6 116
0.060 0.02
0.13 0.85
NE system Van Es 1978
110 106
3.6 56
0.068 0.23
0.03 0.74
NRC 1988 112
111 0.5
43 0.059
0.01 0.07
0.92
R .E. Agnew, T. Yan Livestock Production Science 66 2000 197 –215
209
likely to be the ME . This can be demonstrated in
m
the validation of SCA 1990 which uses the same k
l
as AFRC 1990 but a higher ME by adjusting it
m
with total ME intake. There is no basis to compare the validation between ME and NE systems, since in
the 2 NE systems k is assumed to be same as k to
m l
calculate NEL. However, this assumption is not biologically correct, because research evidence has
indicated that k was higher than k . The difference
m l
between k and k is calculated to be 0.12 or 0.08 in
m l
Van Es 1975 or ARC 1980 when ME GE ratios are from 0.55 to 0.70. The assumption that k
is
m
equal to k in both Van Es 1978 and NRC 1988
l
could thus cover the under-prediction of NE used in
m
these two systems, because the metabolic rates, as discussed previously, are similar between Van Es
1978, NRC 1988 and AFRC 1990. This bias would be larger for late-lactating and dry cows
because the ratios of ME MEI is higher at these
m
two stages. Nevertheless, the more accurate predic- tion of NRC 1988 than Van Es 1978 mainly
reflects that the former adds an activity allowance of 0.10.
6. Conclusions