Combined crossbred and purebred selection for reproduction traits in a broiler dam line

In poultry breeding, the aim is to increase the genetic merit of animals at different stages of the production column (e.g. multiplier and grower). To maximize the genetic response given such an aim, the use of information on both purebred and crossbred performance was proposed: combined crossbred and purebred selection (CCPS; Wei 1992). According to CCPS methodology, purebred and crossbred performance are treated as different but genetically correlated traits. Wei et al . (1991a,b), Baumung (1997b) and Uimari & Gibson (1998) have given a complete theoretical description for this phenomenon. Evidence in chicken can be found in papers by, for example, Orozco & Campo (1975). Wei & Van Der Werf (1994) constructed a simple index combining crossbred and purebred information and showed that under all circumstances CCPS was better than pure line selection (PLS) and crossbred selection (CS). Genetic correlation between purebred and crossbred performance ( rpc ) and crossbred heritability ( h c 2 ) are crucial factors that influenced the relative value of CCPS over PLS & CS. B ijma & van A rendonk (1998) developed a selection index which utilizes both purebred and crossbred information to predict response to BLUP selection with an animal model over multiple generations. It was again concluded that the benefit of including information on crossbred animals was largest when the genetic correlation between purebred and crossbred performance was low. Both Wei & Van Der Werf (1994) and Bijma & Van Arendonk (1998) defined the breeding goal only as an increase in genetic merit of crossbreds, and assumed simplified structures of the breeding programmes, especially in terms of moment of availability of information on performance. Harvey (cited by Hartmann 1992) compared the accuracy of breeding values of purebred animals for crossbred combining ability estimated from different information sources, assuming equal weighting of purebred and crossbred records. Wei & Van Der Werf (1994) suggested that weighting factors of purebred and crossbred information should be derived by the discounted gene flow methodology. As a result, groups of traits (i.e. production versus reproduction, purebred versus crossbred) will have different weightings according to the position of a line in the crossbreeding scheme (Smith 1964; Jiang 1998b). The aim of this study was to construct a breeding goal for a specialized dam line and a realistic scheme for combined purebred and crossbred selection in broilers, and to appraise the relative value of information on purebred and crossbred performance in such a scheme. Materials and methods Consider a three‐way crossbreeding system, A × (C × D), in which A is a specialized sire line, and C and D are specialized dam lines ( Fig. 1). This paper focuses on selection in line D. Generally, both production and reproduction traits should be considered in this line, but here only one reproduction trait is considered, namely, hatching egg number. Discrete generations are assumed. Hens of line D mate to cocks of the same line D to produce purebred offspring. D × D offspring is used for testing and as such includes the selection candidates. Hens of line D also mate to cocks of line C to produce crossbred offspring (C × D) for testing. As a result, every purebred animal has purebred full sibs (FS), purebred half sibs (HS, same sire of line D, different D dams) and crossbred HS (same dam of line D, different sires of line C). After a certain number of offspring are bred, the hens and cocks of line D could be used for other commercial operations, either to produce D × D offspring as great‐grandparent (GGP) stock or to produce C × D offspring as parent stock. 1The position of specialized dam line D in the three‐way crossbreeding system of chickensDefinition of the breeding goal Hatching egg number is expressed at various stages of the production column, for example, in a three‐way crossbreeding system it is expressed by great‐grandparents (GGP), grandparents (GP) and parents ( Fig. 1). The lower is the stage in the production column, the larger is the number of animals. In this study, the GP stage is a representative of purebred expression of hatching egg number and parent stage is a representative of crossbred expression. Index selection methodology (Hazel 1943) is applied. Hatching egg number for purebred and crossbred birds are treated as different but genetically correlated traits. The breeding goal consists of both purebred and crossbred genetic improvements(1) where, H l is the aggregate genotype of a crossbred animal in C × D for selection path l ( l  =  s , selection for sires; l = d , selection for dams), g p [egg/purebred female] and g c [egg/crossbred female] are (true) breeding values for hatching egg number, and v pl and v cl are economic weightings for purebred D and crossbred C*D, respectively, with(2)(3) where ev p (Dfl/egg per purebred female) and ev c (Dfl/egg per crossbred female) are economic values for traits expressed in purebred and crossbred animals; cde pl and cde cl are the number of times purebred and crossbred females, respectively, will express hatching egg number over a time horizon of 20 years (i.e. cumulative discounted expressions for selection path l ). p / c is the ratio of the number of animals at purebred stage to that at crossbred stage. The model by Groen et al . (1998) is used to derive economic values for hatching egg number. In a three‐way crossbreeding system, a GP breeder purchases line D female chicks and line C males, and sells C × D chicks. A parent breeder purchases C × D female chicks and line A males to produce A × (C × D) chicks. The production process of GP is quite similar to that of a parent breeder and cost and revenue components are almost the same. Therefore, we made use of the multiplier part of the model to simulate the production of a GP breeder in a situation of non‐integration. Input parameters are adopted from Jiang et al . (1998a). The parameters that are changed to fit the situation of a GP breeder are in Table 1. Market price of GP line D female chicks (45.00 Dfl/bird) is derived based on a general market reality that the price of GP chicks is about nine to 10 times higher than the price of parent chicks. Market price of C × D chicks (5.00 Dfl/ female) is equivalent to the input price for the parent multiplier stage in Jiang et al . (1998a). The calculated economic values of an improvement of one hatching egg number is 0.7092 Dfl/egg per purebred female at GP stage, and 0.3535 Dfl/egg per crossbred female. 1 . Parameters used in deriving economic values at grandparent (GP) stage, different from parent multiplier stage (based on the model ofGroen et al . 1998)VariableValueMarket price GP line D female chicks (Dfl/bird)45.00Market price parent stock egg hatched (Dfl/egg)2.31Market price parent stock C × D (Dfl/female chick)5.00Number of parent stock eggs per female of line D housed136.00Number of culled eggs per laying GP female9.00The cde of hatching egg number for purebred and crossbred are derived according to the methodology of Jiang et al . (1998b), which is based on the theory of discounted gene flow by Elsen & Mocquot (1974) and Hill (1974). Values of cde are based on an inflation‐free interest rate of 0.0451 and a time horizon of 20 years. The cde are expressed as the number of females at the relevant stage of the production column. Other assumptions required for calculating cde (i.e. generation intervals) are described in the following sections. Results on cde are in Table 2. 2 . Cumulative discounted expressions (cde): the discounted number of future expressions for hatching egg number by purebred and crossbred females in a three‐way crossbreeding of broilers, based on 20 years time horizon and 0.0451 inflation free interest rate, in dependence on generation interval and initial selection pathGeneration intervalSelection pathcdepcdec30.5sire to line D10.4294.862dam to line D10.4295.32532.1sire to line D9.9154.611dam to line D9.9155.07347.3sire to line D6.6453.106dam to line D6.6453.57049.1sire to line D6.3792.986dam to line D6.3793.450Taking into account the difference in number of animals at different stages, the weighting factor for the purebred breeding value is multiplied by p / c (eqn 2). Here, p / c  = 1/[136 (hatching egg) × 0.9 (fertility) × 0.9 (hatchability) × 0.8 (survival) × 0.5 (sex ratio)] = 0.0227. Selection procedure The aim is to appraise the relative values of purebred and crossbred information in a realistic selection scheme. A nucleus of a three‐way broiler crossbreeding structure is assumed. Reproduction levels and a distribution of number of chicks produced by a hen are assumed (Euribrid 1995). A hen starts to produce chicks at 28 weeks of age and ends at 63 weeks of age, in total producing 120 chicks for this first circle of reproduction. A performance record up to 45 weeks of age is the minimum performance record length. There are two alternative selection procedures, resulting in different generation intervals: shortened generation interval (SGI) and normal generation interval (NGI). SGI needs more (testing) capacity than NGI in the rearing period. The following description is based on npo  = 10 (the number of testing/selection candidate female offspring produced by a dam within pure line D up to end of performance testing), nco  = 10 (number of testing female offspring produced by a dam of line D mated to sire of line C up to end of performance testing); alternatives are in Table 3. 3 . Alternatives for family structure*considered in a CCPS selection scheme with either a shortened (SGI) or normal (NGI) generation intervalGeneration interval (weeks)npdnponcoSGINGI550, 5,10, 15, 20, 2530.547.35100, 5,10, 15, 20, 2532.149.11050, 5,10, 15, 20, 2530.547.310100, 5,10, 15, 20, 2532.149.1* npd . number of dams mated to a sire within pure line D; npo , number of testing/selection candidate female offspring produced by a dam within pure line D up to end of performance testing; nco , number of testing female offspring produced by a dam of line D mated to sire of line C up to end of performance testingA schematic representation for SGI is in Fig. 2. Generation i of pure line D has a hierarchical sire/dam family structure. There are ns sires and each sire mates to npd dams; and rx is the sex ratio of purebred selection‐candidates up to the end of performance testing, that is, no. of females/no. of males. Each mating produces npo female and npo / rx male purebred offspring for testing (selection‐candidates). It is assumed that pure‐line breeding starts at the beginning of the reproduction period, that is, at 28 weeks of age. Taking into account mortalities and involuntary culling in different periods, a fixed total survival from day‐old to the end of performance record is used to calculate the required numbers of chicks for testing for all cases. We assume the survival is equal to 80%. A hen can produce the required number (2 × 10/0.8 = 23) of D × D day‐old chicks at the age of 35 weeks. After the required number of purebred chicks is met, hens are mated to cocks of line C, with 1 week of substitution of cocks, and then start at 37 weeks of age to produce nco female testing crossbred offspring. This will finish at 43 weeks of age; testing of D × D is 6 weeks later than the testing of C × D offspring. 2Schematic representation of selection procedure shortened generation interval (SGI). pb = purebreeding, that is, D × D to breed D × D; cb = crossbreeding, that is, D × D to breed C × D. Chicks flow from generation to generation is shown by the dotted lines with arrows ( npo  = 10 = number of testing/selection candidate female offspring produced by a dam within pure line D up to end of performance testing; nco  = 10 = number of testing female offspring produced by a dam of line D mated to sire of line C up to end of performance testing) Testing purebred offspring ( ns  ×  npd  ×  npo / rx males and ns  ×  npd  ×  npo females, generation i  + 1) are retained to 51 weeks of age (during the rearing period, a proportion of male D × D chicks is culled randomly within the family to maintain the required sex ratio rx to the end of performance record). At this time, the crossbred testing offspring C × D is finished for hatching egg production testing up to the age of 45 weeks. Then the best ns purebred sires and ns  ×  npd purebred dams are selected out of ns  ×  npd  ×  npo / rx purebred sire candidates and ns  ×  npd  ×  npo purebred dam candidates for breeding the next generation (generation i  + 2). These selected animals form generation i  + 1. Before they are selected, there has already been pre‐testing of offspring (testing offspring to breed pre‐testing offspring). At the time of selection, these pre‐testing offspring are 51–37 = 14‐weeks‐old for D × D and 8‐weeks‐old for C × D. Among the pre‐testing offspring, ns  ×  npd  ×  npo / rx cockerels and ns  ×  npd  ×  npo pullets, born from the selected ns sires and ns  ×  npd dams of generation i  + 1, are then identified to form the testing offspring (selection candidates of generation i  + 2 and the current generation of testing C × D). This procedure continues from generation to generation. Eggs are hatched the same week as laying (not gathered until end of period and then hatched together). The resulting generation interval is the average age during the replacement time, 30.5 weeks of age. Records for purebred hatching egg number is up to the age of 51 weeks; testing for crossbred performance is up to the age of 45 weeks. The mating sequence in NGI is the reverse of SGI: hens of line D are first assigned to mate to cocks of line C, and then to cocks of line D. Tested D × D from generation i are selected at 45 weeks of age, based on purebred performance up to 45 weeks of age and the current generation crossbred performance (51 weeks of age). The selected ns sires and ns  ×  npd dams are then mated within the pure line to breed the testing D × D as selection candidates of generation i  + 2. This will finish at the age of 53 weeks; the resulting generation interval is also the average age during the replacement time, 49.1 weeks of age. Selection in subsequent generations continues to follow the same procedure. Differences between the alternatives SGI and NGI are in generation interval, testing space for rearing period and performance record length of purebred and crossbred. For SGI, the performance record in purebred is longer than crossbred (e.g. 51 versus 45 weeks of age as from the above example), and for NGI it is the reverse. Selection indices Selection is based on an index combining purebred and crossbred information (Wei & Van Der Werf 1994). The selection index is:(4) where x 1 is crossbred maternal half‐sib sister family phenotypic mean, x 2 purebred paternal half‐sib sister family phenotypic mean (not including the chicken to be selected), x 3 purebred full‐sib sister family phenotypic mean (not including the chicken to be selected), and x 4 is the purebred female individual phenotypic record. b 1  ∼  b 4 are weighting factors for different information sources. As there is no individual phenotypic record for sire candidates, the indices for selection among cocks and hens are distinguished: selection for sires ( I s ) excludes the individual phenotypic record ( x 4 ). Index weights b are calculated according to (matrix notation):(5) where P −1 l is the variance–covariance matrix between information sources x ’s (3 × 3 for I s and 4 × 4 for I d ); G l is the matrix of covariance between information sources of x ’s and breeding values g ’s (3 × 2 for I s and 4 × 2 for I d ); v l includes v pl and v cl (eqns 2 and (3)). The variance of the aggregate genotype for I l is:and the variance of the selection index is:where C is the variance–covariance matrix among traits in the aggregate genotype (2 × 2, the same for selection in sires and dams). Variance due to common environment is assumed to be zero. The genetic gain is:where i s and i d are selection intensities for selection among sires and dams, respectively. Following to gene flow method, cde already include generation intervals. Therefore, Δ G is the genetic gain of one round of selection expressed up to a 20‐years time horizon. For every generation, the number of sires in line D is fixed as 30 ( ns  = 30) and rx is 4. Alternatives for npd , npo and nco are in Table 3. Genetic and phenotypic parameters For egg production in chickens, there are different genetic parameters according to record length (Fairfull & Gowe 1990). In the present study, either 45 or 51 weeks of age are used as record length, for purebred or crossbred animals. Both periods are in the range of the period 2 of Fairfull & Gowe (1990), that is, the middle part record. Thus, a basic heritability for hatching egg number is chosen as 0.23 in this study, although the longer length record tends to be smaller in heritability (Fairfull & Gowe 1990). According to Wei (1992) the heritability of egg production for crossbred is higher than that for purebred. Genetic correlation for this trait between purebred and crossbred is reported in a large scale of change, from 0.10 to more than 0.90 (see Wei 1992). Genetic standard deviation of the traits is 16 eggs (Shalev & Pasternak 1983). Alternative sets of genetic and phenotypic parameters studied are in Table 4. 4 . Alternative sets of genetic parameters for hatching egg number: heritabilities for purebred (hp2) and crossbred performance (hc2), genetic correlation between purebred and crossbred performance (rpc), and genetic standard deviation (σG)( hp 2 )( hc 2 )rpcσ G0.230.250.10−1.0 116 eggs0.230.15−0.45 20.5016 eggs0.15−0.45 20.250.5016 eggs1increment is every 0.10;22  increment is every 0.05Results In this paper various selection procedures, family structures and genetic parameters were studied to appraise the value of CCPS in a broiler dam line in an integrated enterprise. Genetic gains with alternative family structures are derived with an integrated enterprise ( Table 5). For all situations, genetic gain increases with a higher number of crossbred offspring tested per purebred female ( nco ). Information from crossbred animals contributes to genetic gain. This contribution results from an increase in selection accuracy. Increments in genetic gains are shown in Fig. 3 ( Fig. 3a for selection procedure SGI and Fig. 3b for NGI). Here only one situation ( r pc  = 0.3) is shown, but the results for other r pc were similar. The increments in genetic gains were at first sharp, and then levelled of as nco increased. 5 . Genetic gains (ΔG, Dfl/crossbred female) in an integrated operation with alternative family structures*, with selection schemes with shortened (SGI) and normal (NGI) generation intervals and with alternative genetic correlations between purebred and crossbred performance (rpc) (fixed heritabilities for purebred, hp2  = 023, and crossbred, h c 2   = 025, performance, fixed genetic standard deviation (σ G   = 16))Family structurerpc = 0.3rpc = 0.7rpc = 0.9npdnponcoSGINGISGINGISGINGI5504.562.979.246.0611.587.615556.904.5210.336.7812.188.0655107.985.2310.947.1912.718.3455158.625.6611.367.4513.028.5455209.075.9511.637.6313.238.6755259.386.1511.867.7813.398.7851005.913.9011.987.9315.029.9451058.445.5712.998.5815.5810.31510109.656.3813.608.9915.9510.555101510.396.8714.029.2716.2110.725102010.897.2014.319.4616.4110.865102511.247.4314.529.6116.5510.9510505.183.3710.496.8813.158.6110557.755.0711.727.6613.969.16105108.955.8512.428.1314.479.47105159.676.3212.888.2714.809.701052010.146.6413.208.6515.079.861052510.506.8713.448.8015.249.98101006.434.2213.018.5916.3010.77101059.156.0314.149.3316.9711.2110101010.466.9014.819.7717.3911.4810101511.267.4315.2810.0817.7411.6910102011.817.7815.6110.3017.9311.8310102512.188.0415.8710.4718.1111.96* npd , number of dams mated to a sire within pure line D; npo, number of testing/selection candidate female offspring produced by a dam within pure line D up to end of performance testing; nco, number of testing female offspring produced by a dam of line D mated to sire of line C up to end of performance testing3Absolute increment in genetic gain (Dfl/crossbred female) when the number of testing female crossbred C × D offspring per D dam ( nco ) increases, with alternative family structures ( npd / npo denoted as legends; where npo is the number of purebred testing female offspring per dam, and npd is the number of dams mated to a sire within pure line D) in case of integration operation for selection procedure with shortened generation interval (SGI, Fig. 3a) and normal generation interval (NGI, Fig. 3b), with the correlation between purebred and crossbred performance ( r pc ) = 0.3 The npd and npo had a significant influence on the genetic gains ( Table 5). Genetic gains for npd 10/ npo 10 were the highest, npd 5/ nco 10 second, npd 10/ nco 5 third and npd 5/ nco 5 were the lowest. In situation of fixed npd , increasing npo slightly decreased the standard deviation of the aggregate genotype (around 5%) as a higher npo resulted in lower cde values. But increasing npo also led to an increase in accuracy of index and selection intensities in both sire and dam indices. With fixed npo , an increase in npd increased the accuracy of the index and selection intensities for sires (e.g. when rx  = 4, i s was 1.521 for npd 5/ npo 5 and 1.858 for npd 10/ npo 5). Genetic gain for SGI was 52% higher than that for NGI ( Table 5). The higher genetic gains for selection procedure SGI occurred because of a shorter generation interval and thus a higher cde value relative to selection procedure NGI. This led to an increase in standard deviation of the aggregate genotype for SGI. Increasing nco would lead to an increase in the importance of crossbred information in the selection index. Alternatively, nco , as expected, had a significant influence on the b ‐values of sire and dam indices ( Fig. 4). The higher the number of crossbred tested, the higher weighting for crossbred information. When nco  = 0, the percentage relative b 1 ‐value for crossbred to the total b ‐value was, of course, zero. When nco = 5, the value was 20.28% for r pc  = 0.9 and 46.55% for r pc  = 0.3. When nco  = 25, it was 41.60 and 70.28%, respectively. However, the absolute b ‐values for purebred information ( b 2 , b 3 and b 4 ) did not change much with changing nco . 4Ratio (%) of information index value b 1 for crossbred performance ( x 1 ) to the sum of all b ‐value in the index, in relation to alternative number of testing female crossbred C × D offspring per D dam ( nco ) and the genetic correlation between purebred and crossbred performance ( r pc ), with family structure fixed as npd = npo  = 5 ( npo  = number of purebred testing female offspring per dam, and npd  = number of dams mated to a sire within pure line D), for sire index (solid lines) and dam index (dotted lines), in case of integration operation The genetic correlation between purebred and crossbred performance ( r pc ) had an important influence on genetic gain. Higher r pc was associated with higher genetic gain ( Table 6). Again SGI showed a higher genetic gain relative to NGI. The increase in genetic gain for higher r pc , was mainly due to an increase in standard deviation of aggregate genotype and accuracy of the index; accuracy of index for r pc  = 1.0 was almost double that for r pc   = 0.1 (0.60 versus 0.33 for sire index, 0.68 versus 0.33 for dam index). Figure 5 ( Fig. 5a for SGI and Fig. 5b for NGI) shows the relative increment in genetic gains for alternative nco , with different r pc . The lower r pc resulted in a higher relative increment: r pc  = 0.30 resulted in a genetic gain for nco  = 25 that was 107% (sire index) and 105% (dam index) higher than those of nco  = 0. 6 . Standard deviation of aggregate genotype in selection paths for sires and dams (σHs and σHd, Dfl/crossbred female) and genetic gain (ΔG, Dfl/crossbred female) in an integration operation with selection schemes with shortened (SGI) and normal (NGI) generation intervals and with alternative heritabilities for purebred (hp2) and crossbred (hc2) performance and genetic correlations between purebred and crossbred performance (rpc) (fixed genetic standard deviation (σG = 16) and family structure*)Genetic parametersSGINGIhp2hc2r pcσ Hsσ HdΔGσ Hsσ HdΔG0.230.250.113.2214.528.268.519.875.470.230.250.213.3414.648.968.599.955.940.230.250.313.4614.769.688.6610.036.410.230.250.413.5814.8810.538.7410.106.970.230.250.513.7015.0011.538.8110.187.630.230.250.613.8115.1112.538.8910.258.300.230.250.713.9315.2313.568.9610.338.970.230.250.814.0415.3414.859.0410.409.840.230.250.914.1615.4615.919.1110.4810.530.230.251.014.2715.5717.249.1810.5511.410.230.150.510.7711.778.556.937.985.660.230.200.512.3213.4910.137.939.156.710.230.250.513.7015.0011.538.8110.187.630.230.300.514.9416.3612.859.6110.118.100.230.350.516.0817.6213.8410.3511.969.160.230.400.517.1418.7915.0811.0312.769.990.230.450.518.1419.9816.3411.6713.5110.680.150.250.513.5614.8610.778.7310.097.130.200.250.513.6514.9511.238.7810.157.430.250.250.513.7215.0211.548.8310.207.640.300.250.513.7915.0911.998.8810.247.940.350.250.513.8615.1612.188.9210.288.070.400.250.513.9215.2212.508.9610.328.270.450.250.513.9815.2812.689.0010.368.40* Family structure is fixed as nfd  = 5, npo  =  nco  = 10, npo is the number of testing/selection candidate female offspring produced by a dam within pure line D up to end of performance testing; nco is the number of testing female offspring produced by a dam of line D mated to sire of line C up to end of performance testing5Relative increment (%) in genetic gain as the number of testing female crossbred C × D offspring per D dam ( nco ) increases, with alternative genetic correlations between purebred and crossbred performance ( r pc denoted as legends) in case of integration operations for selection procedures with shortened (SGI, Fig. 5a) and normal generation interval (NGI, Fig. 5b). Family structure fixed as npd = npo  = 5 ( npo is the number of purebred testing female offspring per dam, and npd is the number of dams mated to a sire within pure line D) An increase in crossbred and purebred heritabilities has the same impact on genetic gains as an increase in genetic correlation on genetic gains. The increase in genetic gain for higher heritabilities came mainly from an increase in standard deviation of aggregate genotype, while the accuracy of the index only slightly increased with higher heritabilities (not more than 3%). However, crossbred heritability had a stronger influence on the genetic gains than purebred heritability ( Table 6). In fact, the trait with higher economic weighting tended to be more sensitive to the change of its heritability than the trait with a lower economic weighting. Discussion The present study showed that with selection for a combined increase in purebred and crossbred performance (CCPS) including information on crossbred animals always gives additional selection response. However, in different alternatives, the magnitude of the additional response varied much, depending on the genetic background of the population and the perspective taken by the breeders. In previous studies on CCPS (Wei & Van Der Werf 1994; for chicken; Bijma & Van Arendonk 1998; for pigs), the breeding goals were defined only as improved crossbred performance. This holds for some traits in the crossbreeding system of meat‐type animals, e.g. growth and carcass traits, and thus, fits the specialized sire lines. The present study focused on a specialized dam line and assumed a combined selection for both improved purebred and crossbred performance. In specialized dam lines, the importance of reproduction traits at the purebred stage could not be offset by the fact that the number of animals at the purebred stage is much smaller than that of crossbred. Family structures with regard to CCPS In the present study, family structure refers to the sizes and combinations of npd , npo and nco . For all cases, the higher nco would result in higher genetic gains. In practice, nco will be limited in dependence on the reproduction rate of dams and because of organizational reasons (vaccination schedules, utilization of testing facilities etc.) limiting the period to produce progeny from a given mating. In selection for a sire line of chicken with CCPS, Wei & Van Der Werf (1994) found that a higher number of dams mated to a sire favoured CCPS over pure‐line selection (PLS); although it also favoured crossbred selection (CS), responses of CS were always lower than CCPS under given conditions. Bijma & Van Arendonk (1998) also obtained higher response with a higher number of crossbred offspring included in the selection index. However, the efficiency of nco in the role of increasing genetic gains was dependent upon the purebred–crossbred genetic correlation r pc : when r pc was lower, the increment in genetic gain was higher. This is also the case in the study by Wei & Van Der Werf (1994). In the present study, it was found that economic weights for breeding values of purebred and crossbred also had a major influence on the efficiency of increasing nco under CCPS. In the situation of integration, that is, with higher economic weight for crossbred relative to purebred performance, increase in nco gives an appreciate increment in genetic gain. When purebred performance gets a higher economic value relative to crossbred performance (for example in situations of non‐integrated enterprise operation or when the performance level of purebred animals is much lower than the average performance level of other breeding organizations; De Vries 1989), calculations showed that increasing nco only gives marginal increase in genetic gain. Effects of selection procedures The selection procedure with the shortened generation interval (SGI; 30.5–32.1 weeks of age) always resulted in higher genetic gains than the selection procedure with the normal generation interval (NGI; 47.3–49.1 weeks of age). Apart from differences in revenues from SGI and NGI, also differences in costs, for example, in testing capacity, should be considered. With SGI, tested animals produce ‘pre‐testing’ offspring before selection. These pre‐testing offspring have to be housed up to the time of selection of parents, that is, the identification of offspring to be tested (at 14 weeks of age for D × D pre‐testing offspring and 8 weeks of age for C × D pre‐testing offspring). Therefore, the testing capacity required for SNI and NGI will differ. When selection proportion was 10% in dams, the resulted testing capacity of D × D before 14 weeks of age and for C × D before 8 weeks of age, was 10 times for SGI more than for NGI. After that time, the testing capacity was the same for the two selection procedures. Another practical difference between SGI and NGI not considered in the study is the difference in information available on persistency of production at the end of the laying period. SGI assumes records to 45 week of age and NGI assumes records to 51 week of age. A high economic value of persistency of production combined with a high (genetic) variation in this trait would favour the NGI selection procedure. This study focused on maximizing returns (in terms of genetic gain) in dependence on selection procedures with varying testing capacities. In practice, commercial breeding organizations aim for maximizing returns on investment rather than maximizing returns. Moreover, testing capacity may be fixed or limited, and optimization should aim at maximizing returns per testing unit, in other words an optimum ratio of testing purebred and crossbred offspring given the testing capacity. Effects of genetic parameters on the efficiency of CCPS The advantage of CCPS over PLS, and the effect of r pc on the efficiency of CCPS in this study were somewhat different from Wei & Van Der Werf (1994) and Baumung et al . (1997a). Wei & Van Der Werf (1994) suggested that CCPS should replace the commonly used PLS for lower values of r pc . Baumung et al . (1997a) concluded that in cases with a genetic correlation between purebred and crossbred performance close to unity the use of crossbred information in addition to purebred information had a marginal benefit only. Both of them addressed 0.6–0.7 as low r pc , and both defined breeding goal only for crossbred performance. Recent studies showed, that if r pc is higher than 0.7, and even very close to 1.0 (Besbes & Gibson 1998; Merks & Hanenberg 1998). In the present study, the breeding goals were defined as combined increase in purebred and crossbred performance, and varying r pc showed an influence on genetic gains. It is likely that with r pc higher than 0.7, there is no significant increment in genetic gain from an increase in nco . Calculations showed that with a higher economic value for purebred than for crossbred, varing r pc (from 0.3 to 0.9) only had a very small influence on genetic gain. When breeding goals aim at improving crossbred performance only, information on performances of crossbred offspring in addition to available information on purebred relatives, that is, the combined crossbred and purebred selection (CCPS), improves the genetic response. In order to investigate whether or not this also holds when the aim is to increase the performance at both crossbred and purebred stages of the crossbreeding system, the present study constructed a breeding goal and a realistic scheme for the application of CCPS in a specialised dam line of broiler and appraised the relative value of purebred and crossbred performance information in such a realistic scheme. A three‐way crossbreeding system was assumed. The breeding goal was defined based on economic values and cumulative discounted expression of hatching egg number. Two selection procedures were proposed, one with shortened generation interval (SGI, with generation interval of 30.5–32.1 week of age) and the other with normal generation interval (NGI, with generation interval of 47.3–49.1 week of age). Results showed that in such a realistic scheme, increasing the number of crossbred offspring always resulted in extra genetic gain over pure‐line selection. Selection procedure SGI resulted in higher genetic gains, as both absolute values and increment, than NGI, due to shorter generation interval. The effects of family structures and genetic parameters on the efficiency of CCPS were evaluated. It is concluded that including crossbred information gives an increase in genetic gain, but the efficiency of CCPS is depended on selection procedures, family structures and genetic parameters.