Poor sow longevity reduces herd economic efficiency and may cause animal welfare concerns (Stalder et al., 2004). Average annual breeding female replacement rates in US herds have exceeded 60% (PigCHAMP, 1998 to 2004). These rates indicate a substantial economic loss to producers due to increased gilt development costs and reduced pigs per female per day of herd life. Reproductive failure is the most common reason early parity sows are culled (Stalder et al., 2004).
Sow longevity can be increased by improving genetics and environment. Heritability estimates for sow longevity range from 0.05 to 0.27, indicating genetic improvement by selection is possible (Yazdi et al., 2000; Serenius and Stalder, 2004). However, direct selection for sow longevity in nucleus herds is difficult because females are frequently culled before their genetic potential for longevity is expressed. Traits measured on sows before their maximum longevity is reached may be used to indirectly select for sow longevity given a favorable genetic correlation exists within a population. Environment may be the easiest way to improve sow longevity. A whole host of factors including disease, facilities, feeding, and stockmanship influence how long a sow stays in the farm.
Many traits including growth, backfat, age at first farrowing (AFF), and number weaned are associated with sow longevity (Serenius and Stalder, 2004). However, the most important factors have not been quantified under controlled experimental conditions. The purpose of this study was to evaluate the phenotypic association of factors influencing stayability 4 (STAY4), a sow longevity measure defined as the ability of a sow to reach 4th parity, under controlled experimental conditions. Parity 4 was chosen because it was the last parity in which sows were culled using a standardized culling procedure; after 4th parity farmers made the culling decisions.
MATERIALS AND METHODS
Institutional animal care and use committee approval for this project was not required because analyses occurred on an existing data set.
An in-depth description of the data is provided by Moeller et al. (2004). Therefore, an overview will be provided here. The National Pork Producers Council Maternal Line National Genetic Evaluation Program (MLP) was initiated to provide pork producers with unbiased information about differences in reproduction, growth, carcass, and meat quality traits in commercially available maternal lines. Participation was open to all lines that met program definitions of closed populations. Six lines, 1 each submitted by Newsham Hybrid (NH), National Swine Registry (NSR), American Diamond Swine Genetics (ADSG), Danbred (DAN), and 2 lines submitted by Dekalb-Monsanto (DK44 and GPK347), were evaluated. These lines were made by crossing lines maintained by each organization to produce females that expressed 100% maternal heterosis, and thus were considered to be F1 females, consistent with several reports in the literature (Cassady et al., 2002). These lines had mainly Landrace and Large White/Yorkshire origins, but other breeds may have been introduced during their development. For example, the NSR line was an F1 Yorkshire × Landrace crossbred gilt produced by reciprocally crossing Landrace and Yorkshire boars and females at purebred cooperator herds. The 1 unique genetic line, Dekalb-Monsanto GPK347, was produced by inseminating F1 females from the mating of 2 Dekalb-Monsanto maternal lines with semen of boars from the University of Nebraska Index Line. The Nebraska Line was a composite population of Large White and Landrace that was selected for 16 generations for increased ovulation rate, embryonic survival, and litter size at birth (Johnson et al., 1999). Because of the large number of generations separating the Nebraska Index line from the Dekalb-Monsanto lines and the unique selection in its development, the GPK347 line was also considered to be an F1 expressing 100% heterosis.
Entry and removal procedures were standardized in the MLP. Gilts were born in breeder herds. Between 7 and 20 d of age they were transported to 1 of 3 wean-to-finish barns on 3 separate entry dates (February 12, March 5, and April 16, 1997) and entered into the program.
Data recorded at the wean-to-finish barns included entry BW, final BW at 165 d, backfat depth, and LM depth. Backfat and LM depth were measured off the midline at the 10th rib using A-mode ultrasound (600 Series, A-Scan Plus, Sonic Industries, Ithaca, NY). From the wean-to-finish barns, gilts were allocated evenly by genetic line to 1 of 2 new, identical breed-gestation-farrowing units.
Daily estrus detection was practiced and gilts were culled only if they failed to express estrus by 300 d of age or failed to conceive during a 60-d breeding period. Sows were weighed and backfat depth was recorded when they were moved to the farrowing crate and again when their litters were weaned. After litters were weaned, sows were observed for estrus daily and were mated 12 and 24 h after first being observed in estrus. Sows were culled if they failed to conceive within 50 d of weaning their litter or for health reasons identified by a veterinarian. No sows were culled due to poor performance (i.e., litter size, litter weight) through 4 parities. Diets fed to gilts and sows were formulated to meet their requirements. The daily amount of feed allotted to each female during gilt development, breeding, and gestation periods was adjusted based on BW and body condition. Sows were provided ad libitum access to feed until litters were weaned. Amount of feed allotted to each sow during lactation was recorded.
Gilts that never farrowed were excluded from the analysis because they did not have lactation information. Gilt backfat, LM depth, and ADG recorded at the wean-to-finish barns were adjusted to 100 kg of BW according to procedures recommended by the National Swine Improvement Federation (NSIF, 1996). Individual records of prefarrow backfat, number born alive, number weaned, litter weaning weight, lactation feed intake, lactation backfat loss, and lactation length (LL) were adjusted a common parity to remove parity effects using information from parities 1, 2, and 3. This adjustment was calculated using least squares means (PROC GLM; SAS Inst. Inc., Cary, NC) of the previously mentioned traits by parity. Sows with LL less than or equal to 7 d (n = 133, 4%) were deleted from the analysis because they were likely removed for factors not associated with LL (Koketsu and Dial, 1998). Descriptive statistics ± SD of gilt and sow traits recorded on 2,293 females from the MLP are shown in Table 1.
Stayability 4, defined as the ability of a sow to reach 4th parity, was the binary response variable. Therefore, logistic regression analysis (SAS), which is appropriate for binary data, (Johanson and Berger, 2003) was used. The stepwise selection option was used to determine the final model. This option removes nonsignificant effects from the model before an additional effect is added. Each change in the model is listed as a separate step in the output, and at each step a new model is presented. Logistic regression only includes animals in the final model that have records for each trait in the model.
All effects, both categorical and continuous, were nested within genetic line (GL). Effects included in the original model were categorical effects; arrival date to the wean-to-finish unit (entry date) and breed-gestation-farrowing facility (farm), continuous traits recorded on gilts; gilt backfat (GiltBF), LM depth, ADG, age at puberty defined by first observed standing reflex (AGEPUB), and AFF; continuous traits recorded before the last litter of the sow; prefarrow backfat (BFIN), number of piglets born alive (NBA), number of piglets weaned (NW), litter weaning weight (LWW), sow lactation feed intake (LFI), sow lactation backfat loss (BFL), and LL; and quadratic terms were included and allowed to enter model according to the stepwise procedure. The variables LM depth, BFIN, NW, LWW, and BFL were not significant (P < 0.05) in the original model, and thus excluded from the final model by the stepwise procedure. Age at puberty was also excluded. However, when excluding AFF, AGEPUB was significant (P < 0.05) in a similar manner to AFF. The final model contained entry date, farm, GiltBF, ADG, AFF, NBA, LFI, LL, and LL × LL (LL quadratic term), all nested within GL. The mean of each continuous effect (Table 1) was subtracted from each observation to obtain interpretable nonlinear regression coefficients.
Because logistic regression is nonlinear, a coefficient of determination (R2) could not be estimated. Therefore, a generalized coefficient of determination (maximum-rescaled R2) was used to estimate how much of the total variation in the dependent variable (STAY4) was accounted for by the regression function of SAS.
The probability of a sow reaching STAY4 for the continuous traits can be computed from the regression coefficients using the following formula:where exp = exponent, int = intercept, dev_mean = deviation from the trait mean, and reg = the regression coefficient. For example, the probability of a NH female reaching STAY4 with an AFF of 300 d is
RESULTS AND DISCUSSION
Significant factors (P < 0.05) in the final model were entry date, farm, AFF, LFI, ADG, LL, LL × LL (LL quadratic term), GiltBF, and NBA. The maximum-rescaled R2 for the final model was 0.20, indicating the explanatory variables explained 20% of the variation in STAY4. The first factor to enter the final model, AFF, explained 6% of the variation in STAY4. Thus, of the factors studied, AFF had the greatest effect on STAY4.
Odds ratios compare 2 opposing probabilities are used to interpret results from logistic regression. For example, an odds ratio of 1 means that the probability of a sow reaching STAY4 is the same for both effects being compared (entry dates, farms). The odds ratios for GPK347 females at farm 1 and farm 2 were 1.4 and 1.0, respectively (Table 2). This indicates that the GPK347 females had a 40% greater chance of reaching STAY4 in farm 1 than in farm 2, when all other variables were held constant.
Although the conditions between the 2 farms were intended to be identical, farm had a significant effect on STAY4 of the GPK347 sows. The GPK347 sows had greater longevity than other lines (Moeller et al., 2004) but they also were affected more by farm than the other 5 genetic lines (Table 2). Across genetic lines, sows tended to survive better on farm 1 (Table 2). Although farms were structurally the same, perhaps differences between them are explained by feeding management. Sows from farm 1 vs. farm 2 had less (P < 0.05) gestation feed intake (2.20 vs. 2.27 kg) and greater (P < 0.05) lactation feed intake (5.31 vs. 5.13 kg). Although decreased gestation feed intake was favorable for longevity, it appears that some intermediate quantity of intake is optimum as it has also been reported that decreased gestational feed intake is detrimental to longevity (Young et al., 1990).
Entry date was associated (P < 0.05) with STAY4 for the NH, DK44, and DAN genetic lines (Table 2). Explanations why the earliest entry date to the wean-to-finish facilities was generally beneficial are not clear. However, either disease resistance or stockmanship are possible explanations. Often the first group of pigs through a new or clean building will perform better. Likewise, animals that receive more attention may tend to perform better; the first group of gilts likely received more individual attention until groups 2 and 3 arrived.
Significant (P < 0.05) continuous effects recorded on gilts that were associated with STAY4 included AFF, ADG, and GiltBF. Significant (P < 0.05) effects recorded before the last litter of the sow were LFI, NBA, and LL. These associations are described separately below.
Regression coefficients of STAY4 on AFF were significant (P < 0.05) for NH, NSR, ADSG, DK44, GPK347, and DAN females (−0.014, −0.022, −0.017, −0.016, −0.011, and −0.021, respectively; Table 3). When AFF was in the model the effect of AGEPUB on STAY4 was not significant. However, when AFF was removed from the model, AGEPUB was significant in a similar manner to AFF. Therefore, an older AGEPUB or AFF indicates sow reproductive problems later in life.
These findings are in agreement with those of Schukken et al. (1994), Koketsu et al. (1999), and Yazdi et al. (2000) who reported a younger age at first conception or farrowing increased average parity at removal or length of productive life. However, Pomeroy (1960) found that sows with the most parities had an intermediate AFF. Perhaps variation among studies is related to sow BW and ages at sexual maturity because heavier, more mature females at mating consume more lactation feed (Newton and Mahan, 1993) and are less predisposed to farrowing problems (Rozeboom et al., 1996).
Breeding physiologically immature, underweight gilts in a commercial setting is detrimental to lifetime productivity (Williams et al., 2005). However, in this analysis, effects of AFF were estimated while holding other variables constant and a younger AFF indicates greater longevity. Age at puberty or AFF also have a favorable relationship with estrus symptoms (Sterning et al., 1998), the proportion of gilts that farrow a litter (Young, 1998), and nonproductive days (Serenius et al., 2001).
Regression coefficients of STAY4 on ADG were significant (P < 0.05) for ADSG, GPK347, and DAN females (−1.33, −1.42, and −1.37, respectively; Table 3). On average in these lines, stayability decreased as ADG increased, a result consistent with those of Lopez-Serrano et al. (2000) and Tholen et al. (1996), who reported negative (unfavorable) correlations between growth rate and stayability to third parity. However, Stalder et al. (2005) and Yazdi et al. (2000) found no relationship between growth rate and longevity in Landrace sows from US or Swedish populations, respectively. Thus, it appears that ADG may affect sow longevity in some genetic lines while not in others.
Regression coefficients of STAY4 on GiltBF were significant (P < 0.05) for the NH, NSR, and DK44 females (3.86, 4.45, and 6.81, respectively; Table 3). In these lines, STAY4 increased with increasing fatness at 100 kg. These findings are in agreement with Stalder et al. (2005) and Tarrés et al. (2006) who reported 10th-rib backfat <9.0 mm and last rib backfat <16.0 mm, respectively, resulted in decreased lifetime parities and length of productive life, respectively. In contrast, Yazdi et al. (2000) observed no effect of side fat thickness at performance test on longevity in Swedish Landrace. Thus, the effect of GiltBF on sow longevity differs among populations. One might expect this relationship to depend on mean backfat of the population, being more pronounced in leaner lines. However, the relationships within populations are not consistent with line backfat means (Table 1).
The positive effect of backfat on longevity may have occurred because it is unfavorably correlated with fertility traits. Rydhmer et al. (1994) found unfavorable genetic correlations between lean percentage with age at puberty (0.40) and the ability to exhibit the standing reflex (0.10) in Swedish Yorkshire females. Similarly, Serenius and Stalder (2004) reported unfavorable genetic correlations −0.18 and −0.09 between AFF and backfat thickness for Finnish Landrace and Large White, respectively.
The LM depth did not affect (P > 0.05) STAY4 in the present study. However, Stalder et al. (2005) reported fewer lifetime litters in gilts that had less than 36 cm2 of LM area, measured with a real-time ultrasound instrument, than for gilts with larger LM. In the present study, A-mode ultrasound was used to measure GiltBF and LM depth. Perhaps differences in results of the studies are due to different procedures for measuring traits. Moeller (2002) reported that A-mode ultrasound is less accurate than real-time for measuring these traits.
Age at first farrowing and AGEPUB affected (P < 0.05) all genetic lines and had the greatest association with STAY4. Longevity of 3 genetic lines (ADSG, GPK347, and DAN) was affected by ADG, whereas GiltBF affected longevity of the other 3 lines (NH, NSR, and DK44). In general, ADG and GiltBF of gilts during development had less effect on STAY4 compared with AFF and AGEPUB across genetic lines.
Regression coefficients of STAY4 on LFI were significant (P < 0.05) for all lines (ranging from 0.043 for NH to 0.120 for GPK347; Table 3). Greater lactation feed intake resulted in increased probability of reaching STAY4.
Several other studies have found the same relationship between lactation feed intake and sow longevity. Kirkwood et al. (1987) and Prunier et al. (1993) found that reduced lactation feed intake was detrimental to reproductive performance, the primary reason for sow culling. Additionally, lactation feed intake has a more profound effect on the reproductive performance of first-parity sows when compared with multiparous sows (Koketsu et al., 1997).
Regression coefficients of STAY4 on NBA were significant (P < 0.05) for NH, ADSG, and DAN females (0.078, 0.099, and 0.093, respectively; Table 3). Yazdi et al. (2000) also reported that a greater NBA had a positive effect on length of productive life in Swedish Landrace sows. Tholen et al. (1996) reported genetic correlations between number born alive and the ability of a sow to reach 4th parity in 2 Australian pig herds of −0.10 and 0.45. However, in field studies, unlike this study, a favorable relationship between longevity and NBA may be partly because sows with small litters vs. large litters are more likely to be culled from the herd. It is difficult to find a clear explanation for the positive effect of NBA on STAY4. Previously reported literature indirectly supports the current findings. Deen and Xue (1999) reported that sows with more stillbirths had an increased risk of mortality. Irwin and Deen (2000) observed that sow culling increased by 8% when there was at least 1 stillborn pig in the litter and decreased 2% with each extra piglet in the litter. Rozeboom et al. (1996) reported that sows that had farrowing difficulties (which often leads to culling) farrowed fewer live piglets (7.5) than sows that did not have difficulties (9.4). Rozeboom et al. (1996) also found the wean-to-estrus interval after the first litter was longer for sows that had farrowing difficulties than for sows without difficulties (30.5 vs. 12.3 d). Reduced farrowing problems may be an explanation for why an increased NBA improves sow longevity.
Regression coefficients (linear or quadratic) for STAY4 on LL were significant (P < 0.05) for NH, NSR, ADSG, DK44, and DAN females (Table 3). The effect of LL was nonlinear; shorter LL (less than 11 d) were unfavorable for STAY4 for the NH, NSR, ADSG, DK44, and DAN genetic lines, whereas the effect plateaued when LL was 14 d or more. The ability of the productive GPK347 (AFF, AGEPUB; Table 1) to reach the 4th parity was unaffected by LL (Table 3).
Sows with LL less than or equal to 7 d were deleted from the analysis. This was because they were likely removed for factors not associated with LL (Koketsu and Dial, 1998). For example, in the current study, a sow fatality in early lactation would not be related to LL. In a production setting, weaning management may influence LL in relation to stayability. For example, 2 small litters may be combined at birth and 1 sow weaned.
Xue and Dial (1995) categorized LL from 8 to 38 d into short, intermediate, and long groups. They found sows classified with short compared with long LL had a greater chance of being removed from breeding herds. Xue et al. (1997) categorized LL from 12 to 36 d into 5 d intervals. They reported that sows in the shorter LL categories had a greater risk of being removed from the herd compared with sows in the longer LL categories. However, Smith (2006) reported no difference in culling percentage when sows were weaned at 15 or 20 d of lactation. The effect of LL appears to differ in different herds and among different genetic lines.
Number weaned and LWW did not significantly affect STAY4. However, Serenius and Stalder (2004) reported positive genetic correlations of 0.24 and 0.09 between number weaned and length of productive life in Finnish Landrace and Large White populations, respectively. Tholen et al. (1996) observed both positive and negative genetic correlations (−0.09 and 0.38) between 21-d weaning weight and the ability of a sow to reach 4th parity. Differences between studies may be due to different culling strategies. In the current study sows were not culled for poor litter performance (e.g., litter size, litter weight), whereas in other studies this restriction was not applied and producers may have culled sows for poor litter performance.
Prefarrow backfat and BFL did not affect STAY4 in the present study. Previous reports (Tarrés et al., 2006) have suggested an intermediate optimum BFIN (15 to 19 mm) may be favorable for longevity.
Lactation feed intake affected all genetic lines (P < 0.05) and had the greatest association with STAY4 among the lactation traits evaluated. It is evident that practices that increase lactation feed intake within the range evaluated will improve sow longevity.
In conclusion, AGEPUB, AFF, and LFI had the greatest phenotypic relationships with sow longevity under controlled experimental conditions across the genetic lines studied. A younger AGEPUB or AFF under a uniform environment was favorably related to STAY4. Nucleus herds ought to investigate incorporating AFF, or a similar trait, into the breeding objective to reduce sow reproductive problems. Producers should strive to increase LFI through good management (e.g., breeding, facilities, feeding, stockmanship) to enhance sow longevity. The current study did not contain gilts that did not farrow a litter because of the type of analysis conducted. Future research studying the reasons gilts do not farrow is warranted to improve sow longevity.