Genetic Evaluations for Carcass Quality and Feed Efficiency at the Canadian Centre for Swine Improvement (CCSI)

J.R. Brisbane and J. Chesnais
Canadian Centre for Swine Improvement
Ottawa, Canada


As a result of continued selection against backfat thickness, Canadian pigs are getting closer to an optimum level of leanness for domestic and foreign markets. As this occurs, it becomes increasingly important to select on all heritable components of carcass quality, rather than only lean yield. Also, selection must be geared towards the improvement of traits which directly increase the profits of the commercial pork producer.
One of the steps taken toward this goal has been the development of national genetic evaluations for feed efficiency and carcass quality traits. Three national genetic evaluations have been carried out so far (one each calendar quarter), and next year they will replace the genetic evaluation for backfat.

The Genetic Evaluations

The new genetic evaluation system produces estimated breeding values (EBVs) for lean yield, feed conversion, and loin eye area, using a multiple trait animal model. It makes use of a relational database which stores all available pedigree and performance data back to the 1970s. The evaluation system uses measurements on both the live animal and the carcass.

Live-animal measurements include ultrasonic backfat depth, loin muscle depth, and age at 100 kg. We began recording loin muscle depth on a large scale in our home test program in 1996, using the A-mode USM2 probe from the Krautkramer Company, with a 126 mm, 3.5 MHz transducer. The research work on this was carried out by the Québec swine centre (CDPQ). In this work, with a well-trained Krautkramer technician and full cut-outs of 180 pigs, Dion et al. (1998) obtained more accurate predictions of lean yield and loin eye with the Krautkramer USM2 than they did with real-time scans from the Aloka 500.
About 90,000 purebreds are performance tested each year on the swine improvement program, and half of them, including all sire line animals, are now probed for loin muscle depth. Krautkramer no longer manufactures A-mode probes, and it is hoped that real-time probes can eventually be used to get more accurate measurements or information on new traits such as marbling. The US50 real-time probe from Alliance Medical is already being used, and could become the standard in future years.

The carcass measurements included in the evaluation are lean yield, weight of lean in the loin, and loin eye area. These measurements are taken on purebred carcasses in slaughter plants. The genetic evaluation currently includes several thousand of those records.

Data used

Research at CCSI showed that if the backfat data recorded prior to 1990 is not included in the evaluation, the effect on the EBV of current animals is negligible, but there is a large saving in computing requirements. Consequently, the data used in the evaluations comprises:

No feed intake data is used at the moment, but records are being collected and they will be introduced into the system next year. Because there are no feed intake records at the moment, the mixed model equations currently do not include equations for feed intake or feed efficiency breeding values.

Genetic evaluation model

We assume a zero correlation between growth rate and the carcass traits, both genetically and phenotypically, so that EBVs for age at 100 kg are produced using a separate single-trait animal model. These EBVs are then used together with the EBV for carcass traits to produce an EBV for feed conversion ratio which, in the absence of feed intake data, is based on assumed genetic correlations between the traits according to standard BLUP theory. This type of EBV was also used for feed intake traits by Stewart et al. (1990) and described extensively by Harris and Newman (1992).
The model accounts for effects of management group, sex, and measuring technician on the live animal probe traits. Similar effects on the carcass traits are accounted for, as well as a covariate to account for carcass weight effects. The management group is specified by the breeder, with a default grouping by sex within herd-year-quarter of probing if the breeder does not specify a management group.

Heritabilities for backfat, ultrasonic loin muscle depth, loin eye, lean in the loin, and lean yield are assumed to be 52%, 25%, 40%, 40%, and 40%, respectively. A litter variance component is assumed which is 10% of the phenotypic variance for each trait. Genetic correlations with lean yield are assumed to be -0.66, +0.29, +0.65, and +0.78 for backfat, ultrasonic loin muscle depth, loin eye, and loin yield, respectively.

Genetic correlations with feed intake from 25 to 100 kg liveweight are assumed to be +0.43 for age at 100 kg, -0.10 for lean yield, and zero for the other carcass traits.

Some results

Figure 1 shows the average estimated breeding value for lean yield, plotted against year of birth for Hampshire, Yorkshire, Landrace, and Duroc pigs. The EBVs are expressed such that the average for all pigs born from January 1, 1995, to December 31, 1997, within each breed, is zero.

Figure 2 shows the average estimated breeding value for feed conversion ratio (ratio of feed:gain), plotted against year of birth for each breed, and expressed in the same way as in Figure 1.

Figure 3 shows the average estimated breeding value for loin eye area in square centimeters, plotted against year of birth for each breed, and expressed in the same way as in Figure 1.

Annual genetic improvements of -0.5 mm of backfat and -1 day to 100 kg are translating into +0.2% annual improvement in lean yield, -0.014 units annual improvement in feed conversion (about 1.05 kg less feed consumed from 25 to 100 kg liveweight), and +0.3 square centimeters annual improvement in loin eye area.

Summary and Conclusions

Various major enhancements are being made to the Canadian Swine Improvement Program. These are summarized below.

1. Improved selection indexes

Lean yield, feed conversion, and loin eye area are economically important in pork production. New selection indexes have been produced which include these EBVs instead of the backfat EBV which appeared in previous indexes. The new indexes are more relevant to the commercial producer, since they are based on EBV for traits which impact directly on his revenue or costs.

2. More performance test data for improved carcass trait EBV

The widespread collection of loin muscle depth data is increasing the accuracy of genetic evaluations for carcass lean yield. The collection of carcass data from slaughter plants, on a more limited basis, provides sib testing in elite litters which has a large effect on the accuracy of EBV of the best animals.

3. Use of feed intake data for improved feed conversion EBV

Based on the economic values used in the new selection index, and estimated heritabilities and genetic correlations between traits, research (manuscript in preparation) at CCSI has shown that inclusion of feed intake data in genetic evaluations can substantially increase total economic response to selection (by 43% if total feed intake over the grower-finisher period is measured, and by 23% if a feed intake measure is recorded which accounts for 50% of the genetic and phenotypic variation in total intake over the grower-finisher period).

However, individual feed intake is costly to measure, and it is unlikely that large amounts of data will be available from the field. An alternative is to measure the average feed intake of pens, with an appropriate family structure in each pen. Research at CCSI (a manuscript in preparation, results summarized in Appendix 1) shows that with 12 pigs per pen and littermates grouped together, total economic response to selection can be increased by 25%.

Therefore, we plan to use both individual and pen-average feed intake data in the new genetic evaluation system.

Random regression modelling of growth and feed intake, using feed and age data from any given weight (i.e., including young pigs) is currently underway. The random regression approach also allows early genetic evaluations of young pigs and a more accurate selection of which boars to castrate and which to keep entire.


Dion, N., D. Pettigrew, G. Dumas, and J.P. Daigle. 1998. Accuracy of prediction of lean yield, loineye area and marbling from measurements on live pigs. Proc. 6th World Cong. on Genet. Appl. to Livestock Prod. 23:567-570, Univ. of New England, Australia.

Harris, D.L., and S. Newman. 1992. How does genetic evaluation become economic improvement? Proc. of the Symp. on Application of EPD to Livestock Improvement, Pittsburgh, PA. Published by ASAS.

Stewart, T.S., D.H. Bache, D.L. Harris, M.E. Einstein, D.L. Lofgren, and A.P. Schinkel. 1990. A bioeconomic model for swine production: application to developing optimal multitrait selection indexes. J. Anim. Breed. Genet. 107:340.

Appendix 1. Economic Efficiency of Different Types of Feed Intake Recording


The EBV index used is:


The weights are in Canadian dollar units, and the index is designed to measure differences in profit per commercial litter from terminal sires which have the given index values.
The new selection index for sire lines on the Canadian Swine Improvement Program is:

5.94LEAN YIELD - 1.35AGE AT 100KG - 93.75FEED CONVERSION RATIO + 0.40LOINEYE(cm2) (2)

The index (1) used in this study is a simplification of (2) which closely approximates (2), taking into account the index weights in (2) and the genetic correlations between the traits in (2) and the traits in (1). Specifically, the vector b, of index weights in (1) are calculated as:

b = G-1Gca

where G is the genetic covariance matrix for traits in (1), Gc is the matrix of genetic covariances between traits in (1) and traits in (2), and a is the vector of index weights in (2).

The difference between indexes (1) and (2) is that index (1) does not contain loin eye area EBV, and it contains backfat EBV instead of lean yield EBV. Because few animals have direct measurements of carcass loin eye area, and because backfat is highly correlated to lean yield, index (1) is highly correlated (> 0.95 across the hometest population) to index (2).

Realistic genetic parameters are assumed, and selection index methods are used to predict genetic responses, assuming that feed intake in a pen-average recording environment is the same genetic trait as feed intake in a situation where the pig eats from an individual recording machine.


The benefits in Table 1 below are measured as the superiority of the top 10% of pigs on EBV index, in each situation.
For a given cost investment, pen-average recording gives more genetic improvement than individual recording.

For example, if we use pen-average recording with 12 pigs per pen and littermates are penned together, the top 5.5% of pigs selected will have economic superiority of $34.80. This is equivalent to using individual recording and selecting 10% (see table).

Thus, at 12 pigs per pen, if the cost per pig of individual recording is more than twice the cost of pen-average recording, the breeder will get more genetic improvement per unit cost by recording pen averages. The benefit of penning littermates together is very substantial.

Breeders should compare annual depreciation costs on individual intake recording machines with the extra effort required to put ad-lib feeders in each pen and record the amount of feed used by each one.

Table 1. Superiority of the top 10% of pigs on selection index, when
different types of feed intake recording are used
Type of recording
Superiority for backfat (mm)
Superiority for age (days)
Superiority for feed (kg)
Superiority for total $
% advantage over no recording
No recording
Individual (total intake)
Individual (correlation = 0.7 with total)
Average of 12 per pen, littermates apart
Average of 12 per pen, littermates together
Average of 4 per pen, littermates apart
Average of 4 per pen, littermates together