Ewe characteristics associated with neonatal loss in Norwegian sheep

Preventive Veterinary Medicine 114 (2014) 267–275

Contents lists available at ScienceDirect

Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed

Ewe characteristics associated with neonatal loss in Norwegian sheep Ingrid H. Holmøy a,∗ , Steinar Waage a , Yrjö T. Gröhn b a Department of Production Animal Clinical Sciences, Faculty of Veterinary Medicine and Biosciences, Norwegian University of Life Sciences, P.O. Box 8146 Dep. N-0033, Oslo, Norway b Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY 14850, USA

a r t i c l e

i n f o

Article history: Received 21 March 2013 Received in revised form 10 February 2014 Accepted 12 February 2014

Keywords: Case–control study Neonatal loss Recording system Risk factors Sheep

a b s t r a c t A case–control study was conducted to identify ewe characteristics that affect the risk of a ewe losing at least one lamb during the first 5 days post lambing. Data were from a national sheep registry, and only ewes that lambed in the spring of 2010 belonging to flocks that reported disease events were included. Ewes registered with abortion or stillbirth were excluded. Cases (n = 4850) and controls (n = 85,354) from 1153 flocks were studied using logistic regression models, accounting for within flock correlation. The odds of losing at least one lamb increased substantially when litter size exceeded two. For example, in 3year-old ewes, the odds were 6 times greater for those with 3 lambs than for those with 1 lamb. However, the effect of litter size depended on the age of the ewe; for example for ewes giving birth to triplet lambs, the odds of losing at least one lamb were 2.7 times greater in 1-year-old ewes than in 3-year-old ewes. Dystocia was associated with increased risk of losing at least one lamb, but the effect varied by litter size. In ewes with single lambs, the odds of lamb loss were 5 times greater in those that experienced dystocia than in those that did not, while within subgroups of ewes with twins, triplets or >3 lambs, the corresponding odds ratio (OR) of losing one or more lambs was 2.2, 1.5 and 1.3, respectively. Compared with ewes of the Norwegian White breed, ewes of old Norwegian breeds were less likely to lose lambs (OR = 0.8). We also examined the effects of several diseases experienced by the ewe during pregnancy or shortly postpartum on the risk of subsequent neonatal lamb loss. Significantly increased risk was found for ewes with abdominal hernia (OR = 2.5) and for ewes treated for moderate to severe clinical mastitis (OR = 1.6) when compared with ewes without these disorders. In conclusion, our large study population allowed for a detailed analysis of the combined effect of important ewe factors that affected survival of their lambs in the early neonatal period. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Perinatal lamb mortality remains a significant issue to the sheep industry worldwide, and is recognized as the single largest contributor to economic loss in most sheep

∗ Corresponding author. Tel.: +47 22597496; fax: +47 22597083. E-mail address: [email protected] (I.H. Holmøy). http://dx.doi.org/10.1016/j.prevetmed.2014.02.007 0167-5877/© 2014 Elsevier B.V. All rights reserved.

enterprises (Nash et al., 2000; Østerås et al., 2007). The successful rearing of lambs for slaughter and replacement of breeding stock is a key factor in profitable sheep enterprises. Neonatal mortality rates ranging from 6% to14% have been reported from different countries (Wiener et al., 1983; Huffman et al., 1985; Binns et al., 2002). Comparison of figures from the various studies is difficult because rates reported correspond to observation periods of different durations.

268

I.H. Holmøy et al. / Preventive Veterinary Medicine 114 (2014) 267–275

Most neonatal deaths occur during the first days of life (Huffman et al., 1985; Nash et al., 1996; Binns et al., 2002). The exact cause of neonatal deaths is often difficult to determine by post-mortem examination, especially in lambs that die within a few days after birth (Wiener et al., 1983; Green and Morgan, 1993). Complex interactions among prolonged births, low lamb viability, mismothering, starvation and hypothermia are assumed to be main contributors (Wiener et al., 1983; Green and Morgan, 1993; Dwyer and Lawrence, 1998). The relatively high proportion of unexplained early neonatal deaths, even in flocks in which close supervision and adequate intervention are practiced in the lambing season, prompts a closer look at the dam. In Norway, strategies directed towards increased litter size have resulted in increased numbers of triplet and higher order litters. The mean litter size per ewe has increased from 2.0 to 2.2 during the past decade (Holmøy and Waage, unpublished data). Studies have shown decreased neonatal survival rates in lambs from triplet litters compared with lambs from singleton and twin litters (Sawalha et al., 2007; Everett-Hincks and Dodds, 2008). It has been reported that lambs born to 1- or 2-year-old ewes or ewes older than 6 years are at a greater risk of postnatal death than lambs born to 4-year old ewes (Southey et al., 2001; Sawalha et al., 2007). Poorer maternal care is provided by first parity ewes than by multiparous ewes (Dwyer and Lawrence, 2000) and milk yield is lower in first parity ewes than in middle-aged ewes (Ruiz et al., 2000). These factors may affect the capability of ewes to rear large litters, suggesting that the combined effect of ewe age and litter size should be studied more closely. Lambing difficulties decrease the probability of neonatal survival (Miller et al., 2010; Darwish and Ashmawy, 2011). There are great breed differences in the risk of dystocia (Grommers et al., 1985; Dwyer and Bünger, 2012). This may explain parts of the relationship observed between breed and postnatal lamb mortality (Gama et al., 1991; Nash et al., 1996). Diseases occurring during pregnancy may affect the fetus and the viability of the offspring postpartum. Mothering ability of the ewe, in terms of providing resources and care, could be affected by diseases contracted around lambing. Apart from udder diseases (Kirk et al., 1980; Nash et al., 1996; Arsenault et al., 2008), the relationship between periparturient diseases in the ewe and neonatal mortality has not been studied previously. Most studies on perinatal lamb mortality do not distinguish between stillbirths and neonatal deaths. Reports addressing risk factors for neonatal mortality (Huffman et al., 1985; Nash et al., 1996; Christley et al., 2003; Sawalha et al., 2007; Everett-Hincks and Dodds, 2008) were mainly based on repeated observations within one flock or a relatively small number of flocks from certain regions. Thus, more comprehensive studies are needed to estimate, with reasonable precision, the potential effects of various ewelevel risk factors. In Norway, individual ewe data, which include disease records, are available from a national sheep registry. We used data from this registry to identify and quantify the effects of ewe factors that affect the risk of a ewe losing at

least one liveborn lamb during the first 5 days post lambing. Variables studied were age and breed of the ewe, litter size, lambing ease and the calendar time of lambing, and we also examined potential interactions amongst these variables. Further, we aimed to investigate the effect of certain periparturient diseases, a priori suspected to influence neonatal mortality risk. 2. Materials and methods 2.1. Data source We used data from the Norwegian Sheep Recording System (NSRS). One third of all sheep flocks in Norway were enrolled in the NSRS in 2010. In the registry, each animal has a unique identification consisting of numbers for county, municipality within county, flock within municipality, and animal within flock. Recording of several variables is compulsory. One file contains variables recorded for each ewe at lambing and includes the numbers of stillborn lambs and lambs that die before they are given an identity number. Common practice in Norway is to give lambs identity numbers within 24–48 h of birth. Another file contains records of all animals which have received an identity number, where the breed, dates of birth and death of the animals, and the identity number of the dam are available. Disease recording was implemented as a part of the NSRS in 1995. Clinical cases in a flock are recorded on a standardized health card, identifying the particular animal and using specific code numbers for various diseases. Disease events reported to the NSRS are stored in a separate file where the disease episode within a ewe is the unit of observation. Each record contains farm and ewe identification, the date of the event and the disease code number. Regulations regarding traceability of products meant for human consumption oblige all farmers in Norway to keep individual animal records of disease and treatment in their flocks (Ministry of Agriculture and Food, 2005). 2.2. Study design and selection of cases and controls Initially, we extracted from the database records of all ewes that lambed in 2010. We also extracted all records of disease and flock preventive measures in the period from November 1, 2009 to October 31, 2010. Data from the lambing file, the individual animal file and the disease events file were utilized to form one record for each ewe. This primary file included ewes from 3592 flocks. Initial screening and a thorough quality check of the data were undertaken. Identical double records and ewes with no lambs registered were eliminated. Ewes registered with abortion or stillbirth were excluded. Furthermore, we excluded ewes that lambed prior to March 1 or later than June 15. The main lambing season in Norway is in April and May, shortly before pasture is available, and lambing very early or late in the season is rare and considered undesirable. Disease events in one or more ewes were recorded in 1212 flocks, flock preventive measures (e.g. vaccination and deworming), but no disease events, were recorded in 297 flocks, while the remaining 2083 flocks had no such

I.H. Holmøy et al. / Preventive Veterinary Medicine 114 (2014) 267–275

records and were assumed to not report. Because a main issue was to study associations between diseases of the ewe and neonatal mortality we decided to include only those flocks in which disease events had been recorded. In addition, 59 flocks with less than 20 ewes lambing in 2010 were excluded. We used a case–control study design. From the resulting file we selected as cases all ewes that lost at least one lamb during the first 5 days postpartum (n = 4580). The remaining ewes in this file did not experience loss of lambs during this time period and were included as controls (n = 85,354). None of the case or control ewes died within the first 5 days after lambing. Fig. 1 shows the numbers of flocks and ewes that were excluded from the initial data file and gives brief descriptions of the reasons for exclusion. Table 1 shows the distribution of the ewes included in the study by the number of lambs lost and litter size. 2.3. Explanatory variables In Norway, replacement ewes are usually mated when they are around 7 months old and this was practiced in 95% of the flocks registered in the NSRS in 2010. Thus parity equals age in years in most flocks. Litter size equals the total number of lambs born per ewe. Lambing ease is recorded as either “normal”, “assistance, but not dystocia” or “assistance, dystocia”. In the NSRS, there are 25 different code numbers for breed. For analysis purposes, we grouped the breeds into three categories: Norwegian White, Spæl together with other old Norwegian breeds,

269

and all remaining breeds; the breeds contained in the various categories are shown in Table 2. We hypothesized that diseases of the ewe during pregnancy or in the early postpartum period could affect her ability to rear offspring successfully. Diseases recorded between 150 days antepartum and 5 days postpartum were considered as potential explanatory variables. To ensure that the disease preceded lamb death, disease events of a case ewe were included only if they occurred before one of her lambs died. When a lamb died before receiving an identity number, disease events of the ewe recorded after the day of lambing were not considered. If the same disease occurred more than once in the same animal, the date of the first episode was chosen. In the NSRS, cases of clinical mastitis are recorded as either “mild” or “moderate to severe” and these diagnoses were assessed separately. Reduced milk production because of chronic changes of the udder resulting from a previous episode of mastitis could not be ruled out. For ewes older than 1 year we therefore included as an additional variable all cases of clinical mastitis that had occurred during the previous year, i.e., combining records for the two mastitis codes. We also included records of teat injury from the previous year. Diseases that were tested are shown in Table 2. 2.4. Statistical methods Data management and statistical analysis were performed using SAS 9.2 (SAS Institute Inc., Cary, NC, USA). In the initial analysis, age was grouped as 1, 2, 3, 4, 5, 6 and

Fig. 1. Selection of study sample.

270

I.H. Holmøy et al. / Preventive Veterinary Medicine 114 (2014) 267–275

Table 1 Distribution of 89,934 ewes by litter size and number of liveborn lambs lost during the first 5 days post lambing and percentage distribution of the ewes within each litter size. All ewes lambed between March 1 and June 15, 2010 in 1153 Norwegian sheep flocks. Litter size Number of lambs lost 0 1 2 3 >3

1

2

15,395 (97.9) 324 (2.1) -

45,286 (96.7) 1462 (3.1) 109 (0.2) -

3 21,981 (91.5) 1867 (7.8) 148 (0.6) 29 (0.1) -

>3 2,692 (80.8) 485 (14.6) 127 (3.8) 18 (0.5) 11 (0.3)

Total row 85,354 (94.9) 4138 (4.6) 384 (0.4) 47 (0.05) 11 (0.01)

Table 2 Distribution of 89,934 ewes that lambed between March 1 and June 15, 2010 in 1153 Norwegian sheep flocks by various characteristics. Percentage of ewes that lost at least one liveborn lamb during the first 5 days post lambing is given for each variable level. Variable

Level

Number (%) of ewes

Percentage of ewes that lost >0 lambs

Litter size

1 2 3 >3

15,719 (17) 46,857 (52) 24,025 (27) 3333 (4)

2.1 3.4 8.5 19.2

Age of ewe (year)

1 2 3 4 5 6 >6 Missing

21,590 (24) 22,211 (25) 16,632 (19) 12,059 (14) 8354 (10) 5551 (6) 3200 (4) 337

4.4 4.2 4.6 5.8 7.1 7.1 6.7

Lambing ease

Normal Assistance, not dystocia Assistance, dystocia

65,869 (73) 11,503 (13) 12,562 (14)

4.2 5.0 9.8

Breed

Norwegian White old Norwegian breedsa otherb Missing

73,051 (81) 11,661 (13) 5152 (6) 16

5.3 3.8 5.1

Time of lambing

March April 1–15 April 16–30 May 1–15 May 16–31 June

1532 (2) 11,814 (13) 35,612 (39.5) 32,680 (36) 7824 (9) 472 (0.5)

4.4 5.0 5.0 5.3 5.1 4.1

Clinical mastitis, moderate to severe Abdominal hernia Vaginal prolapsec Reproductive systemd Locomotor apparatuse Hypocalcemia Listeriosis Clinical mastitis, mild Uterine prolapse Teat injury Pregnancy toxemia Metritis Mastitis in previous lactationf , g Teat injury in previous lactationf a

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

599 (0.67) 108 (0.12) 150 (0.17) 38 (0.04) 126 (0.14) 177 (0.20) 104 (0.12) 195 (0.22) 68 (0.05) 104 (0.12) 16 (0.02) 801 (0.89) 259 (0.03) 35 (0.004)

10.4 20.4 9.3 13.2 8.7 7.9 8.7 6.7 2.9 5.8 6.3 5.0 4.3 0.0

Includes Spæl, Old Spæl, Furbearing Sheep, Old Norwegian Sheep. Includes Blazed Sheep, Fuglestad Pied, Grey Trønder, Texel, Cheviot, Suffolk, Merino, Finnish Landrace, Blackface, Dorset, Charollais, Romney, and various crossbreeds. c Also including 37 cases apparently recorded erroneously as uterine prolapse prepartum. d Includes cases of induced or delayed parturition and diseases related to the reproductive system at the time of lambing, not covered by other codes. e Includes cases of laminitis, claw problems, paresis, tendon diseases, joint diseases, muscle diseases, pelvic injury, fractures and diseases of the locomotor apparatus, not covered by other codes. f Ewes that were at least 2 years old in 2010 (n = 68,344). g All cases of clinical mastitis. b

I.H. Holmøy et al. / Preventive Veterinary Medicine 114 (2014) 267–275

271

Table 3 Variables significantly associated with the risk of a ewe losing at least one liveborn lamb during the first 5 days post lambing. Multivariable adjusted log odds (ˇ), 95% confidence intervals (CI) and P-values from a generalized linear model with a binomial distribution and a logit link. Generalized estimating equations were applied to account for intra-flock correlation. Analysis included 89,583 ewes that lambed between March 1 and June 15, 2010 in 1153 Norwegian sheep flocks; 4549 ewes lost lambs and 85,034 ewes did not lose lambs. Variable

Levels

Litter size

ˇ

95% CI of ˇ

P

−4.01

Intercept 1 2 3 >3

baseline 0.85 2.28 3.74

0.66 to 1.04 2.04 to 2.51 3.13 to 4.34

1 2 3 4 >4

baseline −0.27 −0.54 −0.10 −0.02

−0.57 to 0.04 −0.97 to −0.10 −0.52 to 0.32 −0.34 to 0.30

0.09 0.02 0.65 0.90

Age of ewe × Litter size

2×2 2×3 2 × >3 3×2 3×3 3 × >3 4×2 4×3 4 × >3 >4 × 2 >4 × 3 >4 × >3

−0.27 −0.50 −0.96 0.02 −0.47 −0.86 −0.38 −0.68 −1.20 −0.14 −0.55 −1.18

−0.60 to 0.06 −0.86 to −0.15 −1.64 to −0.27 −0.44 to 0.49 −0.95 to 0.01 −1.57 to −0.15 −0.85 to 0.09 −1.15 to −0.22 −1.92 to −0.49 −0.49 to 0.21 −0.92 to −0.18 −1.86 to−0.51

0.10 0.005 0.006 0.925 0.052 0.017 0.114 0.003 0.001 0.426 0.004 0.001

Lambing ease

not dystociaa dystocia

baseline 1.62

1.34 to 1.89

<0.0001

Lambing ease × Litter size

dystocia × 2 dystocia × 3 dystocia × >3

−0.83 −1.25 −1.33

−1.11 to −0.53 −1.54 to −0.95 −1.65 to −1.00

<0.0001 <0.0001 <0.0001

Breed

Norwegian White old Norwegian breedsb otherc

baseline −0.21 0.06

−0.36 to −0.07 −0.12 to 0.25

0.004 0.503

0.46 0.93

0.18 to 0.74 0.42 to 1.43

0.001 0.001

Age of ewe (year)

Clinical mastitis, moderate to severe Abdominal hernia

<0.0001 <0.0001 <0.0001

a

Normal and assistance, not dystocia. Includes Spæl, Furbearing Sheep, Old Norwegian Sheep, Old Spæl. c Includes Blazed Sheep, Fuglestad Pied, Grey Trønder, Texel, Cheviot, Suffolk, Merino, Finnish Landrace, Blackface, Dorset, Charollais, Romney, and various crossbreeds. b

>6 years and dummy variables were created using 1-yearold ewes as the reference group. Litter size was grouped as 1, 2, 3 and >3 and analyzed as dummy variables using a litter size of 1 lamb as the reference group. Dummy variables were also generated for the 3 categories of lambing ease, the 3 groups of breed and for the calendar time of lambing, which was divided into 6 time periods (shown in Table 2). The outcome variable was coded as 1 (ewes that lost one or more lambs during the first 5 days postpartum) or nil (ewes that did not lose lambs in this time period). Disease variables were coded as 1 (experienced the disease) or nil (did not experience the disease). Logistic regression was used for testing univariable relationships between each of these explanatory variables and the outcome. Variables that were associated with the outcome with a P-value <0.2 were tested in multivariable logistic regression analysis (SAS Institute, 2012). Ewes within a flock would be expected to be correlated. Generalized estimating equations (GEE) were applied to account for this correlation (Liang et al., 1992),

assuming a compound symmetry correlation matrix, i.e, constant correlation between ewes within a flock. The fixed-effects parameters and variance estimates were estimated in separate procedures and estimates from the model reflect the average effect among the population of animals with a given set of covariates. Stepwise backward elimination was performed until all the variables included were significant at a P-value of ≤0.05 or were considered biologically important. Overall significance of groups of variables, e.g. litter size and age, were tested using Score tests. Biologically plausible interactions between explanatory variables were tested by adding interaction terms to the main effects model. When significant interactions were present, effects were estimated and compared for subgroups defined by combinations of different levels of the interacting variables (a and b). The variance for each combination was calculated using the formula: var (a) + var (b) + 2cov (a, b) (Kleinbaum et al., 1982). The possible presence of multicollinearity among explanatory variables in the final multivariable model was

272

I.H. Holmøy et al. / Preventive Veterinary Medicine 114 (2014) 267–275

examined using weighted predictors (Lesaffre and Marx, 1993). Diagonal values from the information matrix from the final iteration of the multivariable generalized linear model were included as weights in a weighted multivariable regression, using the REG procedure in SAS (Allison, 2012a). Assessment of multicollinearity was based on the variance inflation factors provided by the analysis. One criterion for inclusion in the study was that the ewe belonged to a flock in which disease had been recorded. To examine whether our sample differed from ewes in the source population, distributions of the loss of at least one lamb, litter size, ewe age, lambing ease, and breed were compared for the two groups.

3. Results 3.1. Descriptive findings Five percent of the ewes lost one or more liveborn lambs during the first 5 days post lambing (Table 1). The flocks were located throughout Norway. Mean flock size was 85 lambing ewes (median 72, range 20–632). Crude analysis did not reveal any significant associations between lamb loss and flock size or geographic location of flocks (results not shown). Distribution of the ewes included in the study by various characteristics and the percentage of ewes that lost at least one lamb within each category are shown in Table 2.

3.2. Effects of litter size and age Age effects for ewes >4 years were not different, so age was grouped as 1, 2, 3, 4 and >4 years in the multivariable analysis, using dummy variables. Results from the final multivariable model (Table 3) showed that the risk of a ewe losing at least one lamb increased with increasing litter size. However, the effect of litter size was dependent on the age of the ewe and whether or not assistance at lambing was needed owing to dystocia. Based on estimates from the multivariable model, the odds of a ewe losing at least one lamb were calculated for combinations of age and litter size for the subgroups of ewes with and without dystocia (Table 4). There was a clear tendency that 1-year-old ewes had higher odds of lamb loss than older ewes when litter size was >1 (Table 4A). When grouping 2–4 year old ewes together and comparing these with 1-year-old ewes, the odds of losing lambs, adjusted for lambing ease, breed and significant diseases and accounting for intra-flock correlation, were significantly higher for the latter, both within subgroups of ewes with single lambs (OR = 1.40, 95% CI 1.08–1.80), twins (OR = 1.66, 95% CI 1.46–1.89), triplets (OR = 2.43, 95% CI 2.08–2.84) and >3 lambs (OR = 3.63, 95% CI 2.02–6.53). There was also a tendency that ewes >4 years had a somewhat higher odds of losing lambs than 2–4 years old ewes within all litter sizes (Table 4A). Multivariable analyses comparing these two age groups found that adjusted odds of lamb loss was significantly higher both for ewes with single lambs (OR = 1.27, 95% CI 1.02–1.59), twins

(OR = 1.42, 95% CI 1.22–1.64) and triplets (OR = 1.33, 95% CI 1.20–1.46). 3.3. Effect of lambing ease The proportion of ewes that lost at least one lamb was considerably higher among those with dystocia than in the remaining ewes (Table 2). Univariable estimates for the categories “normal lambing” and “assistance, not dystocia” were not substantially different and these were grouped together as “not dystocia” in the multivariable analysis. There was a highly significant interaction between lambing ease and litter size in the final model (Table 3). The coefficients of the interaction term clearly indicated that the effect of dystocia on lamb loss decreased with increasing litter size. Subgroup analysis showed that the OR (95% CI) for dystocia, adjusted for age, breed and significant disease effects and accounting for intra-flock correlation, was 4.96 (3.78–6.51), 2.18 (1.91–2.50), 1.50 (1.34–1.67) and 1.35 (1.12–1.63) for ewes with singletons, twins, triplets and >3 lambs, respectively. 3.4. Effect of breed The old Norwegian breeds, of which Spæl was the predominant, had a lower risk of lamb loss than Norwegian White (Table 3); the multivariable adjusted OR (95% CI) was 0.81 (0.70–0.93). Interaction terms of breed with litter size, ewe age and lambing ease, respectively, were not significant when added to the final model. Running the final model without the breed variables led to only minor changes in the effect estimates for the other variables in the model. 3.5. Effects of diseases of the ewe Diseases that were examined for possible associations with lamb loss are listed in Table 2. In the multivariable model, two diseases had significant effects. Adjusted odds of lamb loss were 2.53 (95% CI: 1.53–4.17) times greater in a ewe with abdominal hernia than in a ewe without this disorder and 1.58 (95% CI: 1.20–2.09) times greater in a ewe treated for moderate to severe clinical mastitis than in a ewe without a record of moderate to severe clinical mastitis. Ninety-five percent of the cases of moderate to severe clinical mastitis occurred in the period from 10 days before lambing to 5 days after lambing; the remaining cases occurred prior to 10 days prepartum. With the exception of 8 cases, which were recorded post lambing, the cases of abdominal hernia were recorded during the last month prior to lambing. Both diseases occurred most frequently in ewes with litters of multiple lambs; 67% of the ewes with abdominal hernia and 56% of those with moderate to severe clinical mastitis gave birth to litters >2. However, only slight changes in the effect estimates for litter size were observed when the disease terms were removed from the model. The interaction term of moderate to severe clinical mastitis with litter size was not significant when added to the final model. The sparseness of cases of abdominal hernia prevented the interaction term with litter size from

I.H. Holmøy et al. / Preventive Veterinary Medicine 114 (2014) 267–275

273

Table 4 Odds ratios (OR) and 95% confidence intervals (CI) for losing at least one liveborn lamb during the first 5 days post lambing for combinations of ewe age and litter size for (A) ewes without dystocia and (B) ewes with dystocia. The reference is 1-year-old ewes with single lambs that lambed without dystocia. Odds ratios and 95% CI were based on the multivariable adjusted estimates and variance–covariance matrix from the generalized linear model in Table 3. Analysis included 89,583 ewes that lambed between March 1 and June 15, 2010 in 1153 Norwegian sheep flocks; 4549 ewes lost lambs and 85,034 ewes did not lose lambs. Age

Litter size 1 OR

A

B

2

3

95% CI

OR

95% CI

95% CI

OR

95% CI

1 2 3 4 >4

1 0.77 0.58 0.91 0.98

0.56–1.05 0.37–0.92 0.59–1.40 0.70–1.36

2.34 1.36 1.40 1.46 1.99

1.94–2.83 0.89–2.09 0.81–2.41 0.87–2.45 1.29–3.08

9.75 4.52 3.56 4.46 5.52

7.75–12.3 2.89–7.08 2.04–6.22 2.61–7.64 3.50–8.70

42.1 12.4 10.4 11.4 12.6

23.5–75.1 6.16–24.8 4.79–22.4 5.37–24.4 6.27–25.5

1 2 3 4 >4

5.03 3.85 2.94 4.56 4.93

3.88–6.52 0.67–5.80 1.74–4.98 2.75–7.57 3.24–7.49

5.17 3.01 3.09 3.21 4.39

3.57–7.48 2.17–5.13 1.65–5.79 1.74–5.91 2.56–7.53

14.12 6.55 5.16 6.46 7.98

9.53–20.9 3.78–11.3 2.71–9.80 3.46–12.1 4.58–13.9

55.8 16.4 13.8 15.2 16.8

28.9–108 7.65–35.3 6.05–31.3 6.70–34.5 7.78–36.2

being tested. Removing lambing ease from the final model led to only a slight increase in the point estimate of the OR for abdominal hernia (OR = 2.8). Most ewes in Norway that suffer an episode of mastitis or experience injury or abnormalities of the udder or teat during lactation are culled after weaning, and only 294 of the 68,344 ewes older than 1 year in our study (0.04%) had a record of clinical mastitis or teat injury in the preceding lactation. Multivariable analysis of the subset of ewes >1 year of age, adjusting for covariates included in the final model for the complete dataset (Table 3), did not find any significant effects of a previous record of mastitis or teat injury on the risk of losing lambs.

3.6. Multicollinearity As would be expected, variables coded as dummy variables and included in interaction terms, i.e., age, litter size and lambing ease, were correlated with the interaction terms. Variance inflation factors for the dummy variables included in interaction terms ranged between 4.6 and 39.8. Multicollinearity could affect the precision of the estimated model parameters by inflating the standard errors for the parameters. However, confidence intervals of main effects and interaction terms of these variables were not very wide (Table 3). Multicollinearity can be ignored if the high variance inflation factors are caused by the inclusion of powers or products of main effect variables in the model (Allison, 2012b). For variables included in the final model as main effects only, the variance inflation factors were <1.02.

3.7. Study sample compared with source population With the exception of lambing ease, the distribution patterns within the study sample of the ewe characteristics included in the study were similar to the patterns in the source population defined in Fig. 1 (data not shown). Dystocia was recorded in 11% of the ewes in the source population and in 14% of the ewes in our sample. Flock size was 85 and 80 and the proportion of ewes that lost at

OR

>3

least one lamb was 5.1% and 4.8% in the sample and source population, respectively. 4. Discussion This population-based case–control study identified and quantified the relationships between several ewe characteristics and the risk of losing at least one lamb in the early neonatal period. The large number of ewes included in our study made it possible to examine in detail whether interactions between explanatory variables were present. Separate effects of litter size, ewe age and lambing ease have been reported previously; however, the complex interactions between these factors have not been considered in detail. In accordance with previous studies (Southey et al., 2001; Sawalha et al., 2007; Everett-Hincks and Dodds, 2008; Hatcher et al., 2009), we found that the total number of lambs born to the ewe was a strong risk factor. The effect of litter size depended to a large extent on the age of the ewe and whether or not she had experienced dystocia. Among ewes that did not experience dystocia, the odds of losing at least one lamb were, within all age groups, around twice as high for ewes with twins as compared with those with single lambs. At the lamb level, this corresponds to a similar mortality risk of single- and twin-born lambs. When litter size exceeded two, the mortality risk increased markedly. Everett-Hincks and Dodds (2008), who analyzed observations at the lamb level, found that the survival rate to 3 days of age was 8% greater for twins than for triplets but did not differ between singletons and twins. In another study, also using lamb as the unit of observation, it was estimated that death during the first week of life occurred much more frequently in lambs in triplet and higher order litters (36%) than in twins (19%) and single lambs (8%) (Hatcher et al., 2009). These studies did not include lambs born to 1-year-old ewes and mortality rates reported did not differentiate between ewes with and without dystocia. Studies that observed lambs during an extended postnatal period, also found that mortality risk increased markedly with increasing litter size (Southey

274

I.H. Holmøy et al. / Preventive Veterinary Medicine 114 (2014) 267–275

et al., 2001; Sawalha et al., 2007). In triplet and larger litters in our study, there was usually only one lamb that died, demonstrating that exposures or characteristics associated with mortality mainly were at the individual lamb level and not at the litter level. When dystocia occurred, the risk of lamb loss increased considerably in ewes with twins, but far more in those with single lambs. Previous studies also found that dystocia was a more frequent cause of death in single lambs than in lambs in larger litters (Dalton et al., 1980; Hall et al., 1995). Average birth weight is larger in single-born lambs than in twins (Hatcher et al., 2009) and fetopelvic disproportion is recognized as a major cause of dystocia in ewes with single lambs (Grommers et al., 1985; Arthur and Bee, 1996; Speijers et al., 2010). In cases of dystocia associated with litters of multiples, often due to malpresentation of lambs (Speijers et al., 2010), successful obstetric delivery is achieved with less difficulty because of the smaller fetuses. In Norway, ewes usually lamb for the first time when they are 1-year-old, and for all litter size and lambing ease combinations, the neonatal mortality risk was considerably higher for lambs born to 1-year-old ewes than for lambs born to older ewes. In cases of triplet litters and, in particular, higher order births, the risk was extremely high. Previous studies have addressed the relationship between age of the ewe and perinatal lamb mortality; however, early neonatal loss of lambs for 1-year-old ewes compared with that for older ewes has not been investigated specifically. Intake of sufficient amounts of milk, including colostrum, is critical for lamb survival. Milk production is lower in first parity ewes than in older ewes (Ruiz et al., 2000) and inadequate supply may occur in ewes with multiple lambs. Mothering ability is also essential for survival and first parity ewes are more likely to express withdrawal, aggression and lack of co-operation with newborn lambs’ sucking attempts (Dwyer and Lawrence, 2000). In line with a previous study of flock level neonatal mortality in Norway (Holmøy et al., 2012), we found that ewes of old Norwegian breeds, including mainly the Spæl breed, had somewhat lower odds of losing at least one lamb than ewes of Norwegian White. Breed differences in postnatal and neonatal mortality risks have also been reported from other countries (Wiener et al., 1983; Gama et al., 1991; Nash et al., 1996). Ewes with clinical mastitis during pregnancy or in the early postpartum period were at increased risk of losing lambs. Few studies have examined the relationship between ovine mastitis and neonatal lamb survival. Arsenault et al. (2008) found a five-fold increase in the hazard of postnatal lamb mortality for ewes with clinical mastitis, whereas Kirk et al. (1980) did not find any significant association between clinical or subclinical staphylococcal mastitis and postnatal lamb survival. Mastitis is associated with reduced milk production and quality, which may result in inadequate milk supply to the lambs, particularly in ewes with multiple lambs. Consequences of abdominal hernia in the ewe on lamb survival have not been reported previously. Access to the udder can be difficult in severe cases of hernia, which may partly explain the increased risk of neonatal lamb loss observed in our study.

Data quality is essential when using registry data. In general, members of the NSRS are highly motivated to ensure that data are recorded completely and correctly. Membership is voluntary and records are used for breeding and management purposes at the flock level and for the selection of rams for breeding at the local and national level. Only flocks with disease records were included, but there might be some doubt whether or not disease events were recorded completely. For most of the diseases that were tested, the diagnosis was straightforward and misclassification of cases using an incorrect diagnosis code was likely not an important problem. If disease events were not recorded and thus misclassified as non-events, the strength of a real effect would be underestimated in the analysis. Therefore, our estimates for the effects of mastitis and abdominal hernia are likely conservative. Furthermore, the low frequencies of the diseases in the study sample imply a risk of committing type II errors, i.e., not detecting true effects. In some cases, a liveborn lamb that died very shortly after birth might have been registered as a stillborn lamb. This would hardly affect our analysis because ewes with stillborn lambs were not included among either the cases or controls and major differential patterns of the explanatory variables between the cases included in the study and those that might have been erroneously registered as stillborn were not very likely. Given the strong influences of management practices and potential environmental exposures on the risk of neonatal death (Fogarty et al., 1992; Everett-Hincks and Dodds, 2008; Holmøy et al., 2012), flock was likely an important confounding factor and was therefore included as a random factor in the analysis. Our large study population represented the Norwegian sheep population fairly well. Common characteristics of almost all flocks in Norway are that ewes are kept indoors during pregnancy, lambing takes place indoors between the end of March and the end of May and lambing ewes and newborn lambs are monitored quite closely. A detailed description of housing and management of sheep in Norway has been presented recently (Holmøy et al., 2012). Some exposure to adverse weather conditions may occur; however, such factors are less important risk factors for neonatal lamb mortality compared to what has been reported from other countries (Fogarty et al., 1992; EverettHincks and Dodds, 2008). Thus, among the risk factors for neonatal mortality, the relative importance of ewe characteristics likely is greater in our study population than in sheep enterprises where lambing occurs outdoors and close monitoring of individual animals is difficult. 5. Conclusions The risk of a ewe losing one or more lambs in the early neonatal period was strongly affected by combined effects of litter size, age of the ewe and whether or not she experienced dystocia. One-year-old ewes that gave birth to more than two lambs and experienced dystocia were at a particularly high risk of lamb loss. Also, ewes with a record of moderate to severe clinical mastitis or abdominal hernia were at increased risk of losing at least one lamb, while

I.H. Holmøy et al. / Preventive Veterinary Medicine 114 (2014) 267–275

the occurrence of other common diseases in the ewe in the periparturient period was not found to affect neonatal lamb mortality. Conflicts of interest There are no conflicts of interest. Acknowledgements Access to data was provided by the Norwegian Sheep Health Service, Animalia, Oslo. The Agricultural Agreement Research Fund, the Foundation for Research Levy on Agricultural Products, Animalia–Norwegian Meat and Poultry Research Centre and the Norwegian Sheep and Goat Association are acknowledged for funding this study. The Committee for Research and Ethics at the Norwegian School of Veterinary Science is acknowledged for funding a research stay for the first author at the Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University. References Allison, P.D., 2012a. Logistic Regression Using SAS® : Theory and Application, 2nd ed. SAS Institute Inc., Cary, NC, pp. 60–63. Allison, P.D., 2012b. When can you safely ignore multicollinearity? http://www.statisticalhorizons.com/multicollinearity. [Accessed Jan. 7, 2014]. Arsenault, J., Dubreuil, P., Higgins, R., Bélanger, D., 2008. Risk factors and impacts of clinical and subclinical mastitis in commercial meat-producing sheep flocks in Quebec, Canada. Prev. Vet. Med. 87, 373–393. Arthur, G.H., Bee, D., 1996. Dystocia and other disorders associated with parturition. General considerations. In: Arthur, G.H., Noakes, D.E., Pearson, H., Parkinson, T.J. (Eds.), Veterinary Reproduction and Obstetrics. , 7 ed. Saunders, London, pp. 185–194. Binns, S.H., Cox, I.J., Rizvi, S., Green, L.E., 2002. Risk factors for lamb mortality on UK sheep farms. Prev. Vet. Med. 52, 287–303. Christley, R.M., Morgan, K.L., Parkin, T.D.H., French, N.P., 2003. Factors related to the risk of neonatal mortality, birth-weight and serum immunoglobulin concentration in lambs in the UK. Prev. Vet. Med. 57, 209–226. Dalton, D.C., Knight, T.W., Johnson, D.L., 1980. Lamb survival in sheep breeds on New Zealand hill country. N. Z. J. Agric. Res. 23, 167–173. Darwish, R.A., Ashmawy, T.A.M., 2011. The impact of lambing stress on post-parturient behaviour of sheep with consequences on neonatal homeothermy and survival. Theriogenology 76, 999–1005. Dwyer, C.M., Lawrence, A.B., 1998. Variability in the expression of maternal behaviour in primiparous sheep: effects of genotype and litter size. Appl. Anim. Behav. Sci. 58, 311–330. Dwyer, C.M., Lawrence, A.B., 2000. Maternal behaviour in domestic sheep (Ovies aries): constancy and change with maternal experience. Behaviour 137, 1391–1413. Dwyer, C.M., Bünger, L., 2012. Factors affecting dystocia and offspring vigour in different sheep genotypes. Prev. Vet. Med. 103, 257–264. Everett-Hincks, J.M., Dodds, K.G., 2008. Management of maternaloffspring behavior to improve lamb survival in easy care sheep systems. J. Anim. Sci. 86, E259–E270. Fogarty, N.M., Hall, D.G., Holst, P.J., 1992. The effect of nutrition in mid pregnancy and ewe liveweight change on birth weight and management for lamb survival in highly fecund ewes. Aust. J. Exp. Agric. 32, 1–10.

275

Gama, L.T., Dickerson, G.E., Young, L.D., Leymaster, K.A., 1991. Effects of breed, heterosis, age of dam, litter size, and birth weight on lamb mortality. J. Anim. Sci. 69, 2727–2743. Green, L.E., Morgan, K.L., 1993. Mortality in early born, housed lambs in south-west England. Prev. Vet. Med. 17, 251–261. Grommers, F.J., Elving, L., Van Eldik, P., 1985. Parturition difficulties in sheep. Anim. Reprod. Sci. 9, 365–374. Hall, D.G., Fogarty, N.M., Gilmour, A.R., 1995. Performance of crossbred progeny of Trangie Fertility Merino and Booroola Merino rams and Poll Dorset ewes. 1. Lamb birth weight, survival and growth. Aust. J. Exp. Agric. 35, 1069–1074. Hatcher, S., Atkins, K.D., Safari, E., 2009. Phenotypic aspects of lamb survival in Australian Merino sheep. J. Anim. Sci. 87, 2781–2790. Holmøy, I.H., Kielland, C., Marie Stubsjøen, S., Hektoen, L., Waage, S., 2012. Housing conditions and management practices associated with neonatal lamb mortality in sheep flocks in Norway. Prev. Vet. Med. 107, 231–241. Huffman, E.M., Kirk, J.H., Pappaioanou, M., 1985. Factors associated with neonatal lamb mortality. Theriogenology 24, 163–171. Kirk, J.H., Huffman, E.M., Anderson, B.C., 1980. Mastitis and udder abnormalities as related to neonatal lamb mortality in shed-lambed range ewes. J. Anim. Sci. 50, 610–616. Kleinbaum, D.G., Kupper, L.L., Morgenstern, H., 1982. Epidemiologic Research. Principles and Quantitative Methods. JohnWiley & Sons, Belmont, pp. 485–487. Lesaffre, E., Marx, B.D., 1993. Collinearity in generalized linear regression. Commun. Stat. Theory Methods 22, 1933–1952. Liang, K.-Y., Zeger, S.L., Qaqish, B., 1992. Multivariate regression analyses for categorical data. J. R. Stat. Soc. Series B. Stat. Methodol. 54, 3–40. Miller, D.R., Blache, D., Jackson, R.B., Downie, E.F., Roche, J.R., 2010. Metabolic maturity at birth and neonate lamb survival: association among maternal factors, litter size, lamb birth weight, and plasma metabolic and endocrine factors on survival and behavior. J. Anim. Sci. 88, 581–593. Ministry of Agriculture and Food, 2005. FOR 2005-11-13 nr 1356: Forskrift om merking, registrering og rapportering av småfe § 9. Available online: http://www.lovdata.no/for/sf/ld/td-200511301356-0.html (In Norwegian) [Title translation: Regulation on identification, registration and reporting of small ruminants.]. [Accessed Mar. 1, 2013]. Nash, M.L., Hungerford, L.L., Nash, T.G., 2000. Evaluating economic consequences of risk factor based health improvement decisions. In: Proceedings of the 9th Symposium of the International Society for Veterinary Epidemiology and Economics, Breckenridge, Colorado, p. 431. Nash, M.L., Hungerford, L.L., Nash, T.G., Zinn, G.M., 1996. Risk factors for perinatal and postnatal mortality in lambs. Vet. Rec. 139, 64–67. Østerås, O., Gjestvang, M.S., Vatn, S., Sølverød, L., 2007. Perinatal death in production animals in the Nordic countries – incidence and costs. Acta Vet. Scand. 49, S14. Ruiz, R., Oregui, L.M., Herrero, M., 2000. Comparison of models for describing the lactation curve of Latxa sheep and an analysis of factors affecting milk yield. J. Dairy Sci. 83, 2709–2719. SAS Institute, 2012. SAS 9.2 Online Documentation http://support.sas. com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm #statug genmod sect002.htm [Accessed Mar. 1, 2013]. Sawalha, R.M., Conington, J., Brotherstone, S., Villanueva, B., 2007. Analyses of lamb survival of Scottish Blackface sheep. Animal 1, 151–157. Southey, B.R., Rodriguez-Zas, S.L., Leymaster, K.A., 2001. Survival analysis of lamb mortality in a terminal sire composite population. J. Anim. Sci. 79, 2298–2306. Speijers, M.H.M., Carson, A.F., Dawson, L.E.R., Irwin, D., Gordon, A.W., 2010. Effects of sire breed on ewe dystocia, lamb survival and weaned lamb output in hill sheep systems. Animal 4, 486–496. Wiener, G., Woolliams, C., Macleod, N.S.M., 1983. The effects of breed, breeding system and other factors on lamb mortality: 1. Causes of death and effects on the incidence of losses. J. Agric. Sci. 100, 539–551.