| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* EpiCentre, Institute of Veterinary, Animal and Biomedical Sciences, Massey University, New Zealand
Faculty of Veterinary Sciences, University of Dublin, Ireland
1 Corresponding author: c.heuer{at}massey.ac.nz
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Key Words: bulk tank antibodies against bovine viral diarrhea virus • economic effects • herd-level analysis • partial budget
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
An initial estimation of BVDV infection status of the herd can be made using a bulk tank milk (BTM) sample (Bitsch and Rønsholt, 1995). Good correlation exists between seroprevalence and BTM antibody concentrations (Niskanen, 1993). A strong association between seroprevalence in the herd and the likelihood of a PI being present has also been demonstrated (Houe and Meyling, 1991). To relate BTM inhibition percentages (%INH) to the likelihood of the presence of PI animals currently in the herd, spot testing of young stock at 6 to 18 mo of age (after maternal antibodies have waned) provided a reliable indication of current infection and the likely presence of a PI animal in the herd (Houe and Palfi, 1993; Houe, 1994; Bitsch and Rønsholt, 1995). After sampling a subset of young stock in dairy herds, receiver operator characteristic (ROC) analysis suggested an optimal cut-off for BTM of 80%INH, giving 81.2% sensitivity and 91.2% specificity for the detection of PI animals in a New Zealand study (Thobokwe et al., 2004).
Several authors have examined reproductive performance in association with BVDV status either at the individual animal level using serum antibody status (Houe et al., 1993a; Larsson et al., 1994; Rüfenacht et al., 2001), or at the herd level using 2 or more BTM samples at intervals of 4 to 18 mo to determine the herd BVDV status (Fredriksen et al., 1998; Valle et al., 2001; Robert et al., 2004). With the exception of an increased risk of abortion, which was most frequently associated with BVDV infection (Larsson et al., 1994; Fredriksen et al., 1998; Rüfenacht et al., 2001), the studies mentioned have given rise to conflicting results. This is not surprising considering the differences in study design, power, classification of BVDV status, and method of analysis. Although a decreased conception rate has been reported (Houe et al., 1993a; Larsson et al., 1994), other studies failed to demonstrate negative effects of BVDV on return rates, number of inseminations per cow, and calving intervals (Fredriksen et al., 1998; Rüfenacht et al., 2001; Valle et al., 2001). Robert et al. (2004) reported an increased risk of late returns (>25 d) in BVDV-infected herds, but did not observe an association between the rate of 3-wk returns and herd BVDV status.
The primary aim of this study was to identify potential reproductive and production loss in BVDV-infected herds in New Zealand, as determined by a single BTM sample. This study also aimed to explore the association between BVDV status and cull rate related to infertility, for which data are lacking. Most of the epidemiological studies previously referred to in this introduction were based in Europe. The small size of European study herds (12 to 26 cows) (Rüfenacht et al., 2001; Valle et al., 2001; Robert et al., 2004) and their management (housing, concentrate feeding, less strict seasonal calving) is considerably different from the average dairy herd in New Zealand. The impact of BVDV infection might therefore vary, as might the ability to detect its effects in herds.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
80%INH BVDV antibody in BTM indicating herd infection status. For the current study, information on individual reproductive and milk production performance of cows was available from the Livestock Improvement Corporation (LIC) database (LIC, Private Bag 3016, Hamilton, New Zealand) and was compared with that cut-off point.
Data Validation
Available herd information from 632 herds (267,171 cows) was downloaded from the LIC database and validated using Access (Microsoft Corp., Redmond, WA). In total, 2,316 cows from 3 herds were initially excluded from the database for the following reasons: 84 cows had duplicate entries (same lifetime identification and same herd code) but were listed with a different exit reason; a further 1,640 cows were allocated to 2 different herds with 2 different results from the bulk tank milk testing; and 592 animals were allocated to 2 different herds at the same time with different BTM test results, and thus were regarded as entry errors.
Only animals with calving dates from June 1, 2001, to March 31, 2003 (590 herds, 185,050 animals) and mating dates from September 1, 2001 to May 31, 2002 (587 herds, 146,038 animals) were included in this study. These periods included the date of BVDV BTM antibody test in March 2002.
Response Variables
The following variables of interest were calculated as herd-level averages during the 2001–2002 season: annual pregnancy rate, first-service conception rate, calving-to-conception interval, annual abortion rate, overall annual culling rate, annual culling rate because of infertility, proportion physiological service intervals, and milk solids produced per day adjusted for lactation stage and parity.
Pregnancy Rate, First-Service Conception Rate, Calving-to-Conception Interval, Abortion Rate, Induction Rate, and Culling Rate.
The calculation of the performance indicators required cows with complete records of calving and mating dates in 2001–2002 as well as calving or removal dates of the subsequent season, 2002–2003. There were 571/590 herds (96.8%) left with complete information to calculate pregnancy rate, first-service conception rate, and calving-to-conception interval, 570/590 herds (96.6%) to calculate culling rates, and 565/590 herds (95.8%) for abortion and induction rates.
The gestation period was calculated as the difference between the last mating date in the 2001–2002 season and the calving date in the subsequent season. A key was derived that defined the pregnancy status (yes/no), first-service conception (yes/no), and the time between calving and conception for each animal. If no information about the pregnancy outcome was stated in the LIC database, a gestation period of 270 to300 d was considered as "normal," a gestation period of 220 to 269 d as "induced," and a gestation period of 63 to 219 d as "abortion."
Cows with those gestation periods were assumed to have conceived after first mating if they only had one mating in that season; otherwise, first-service conception was assumed unsuccessful. In addition to cows with normal gestation periods, a successful first-service conception was recorded for animals in which an induction or abortion was stated in the database and that had a gestation period of <330 d or <250 d, respectively, and only one service. Animals without information about calving dates of the subsequent season were considered nonpregnant if they were culled or sold because of infertility. If any other reason than infertility was stated in the records of such cows, they were excluded from the analysis of first-service conception rates.
The calving-to-conception interval was calculated as the difference between the calving date and the last mating date in the season 2001–2002. For animals with gestation periods longer than 300 d and no stated pregnancy outcome, we assumed another mating date 21 d after the recorded mating and calculated the calving-to-conception interval using that date.
Proportion Physiological Service Intervals.
Only the interval between the first and second service was considered because later intervals were regarded as unreliable because of possible hormonal treatment or interference with natural services in the later part of the mating period. Cows without repeated service were excluded from this analysis. Finally, 458/590 herds (79.0%) had complete herd and mating information to calculate their proportion physiological service intervals.
Service intervals of 18 to 24 and 38 to46 d between the first 2 services were defined as physiological, intervals of 25 to 37 or >46 d were defined as nonphysiological. Intervals <18 d were excluded from the analysis because of the possibility of hormonal treatment or repeated AI within the same, possibly extended, estrus event.
Milk Production and SCC.
There were 541/590 (91.7%) herds with complete information on milk test data performed from June 2001 through June 2002. A lactation model was developed to compute a milk production parameter that was comparable between herds using Wood’s equation (Wood, 1976). The model standardized milk production on 4.5% milk fat and 3.2% protein. The model equation was
 |
where FPCM = milk volume x percentage fat/4.5 x percentage protein/3.2; a, b, c, and d were regression coefficients calculated for each herd; ln = natural logarithm; and age was years from birth to test date. Estimated mean values were generated for each herd for a 3-yr-old cow being 100 d in milk (age = 3, DIM = 100). One herd was excluded because of an unlikely high mean value.
Individual SCC were transformed by natural logarithm and averaged at the herd level.
Statistical Analysis
Herd BVDV antibody (%INH) was the main risk variable of interest. For categorical analysis, it was converted to an ordinal scale with 
40%INH classified as BVDV antibody class 1, 41 to 50%INH as class 2, 51 to 60%INH as class 3, 61 to 70%INH as class 4, 71 to 80%INH as class 5, and >80%INH as class 6. Mean, standard deviation, median, and 95% confidence intervals were used to describe the response variables.
The relationship between the response of interest and herd BVDV antibody %INH was assessed using multivariate regression models. Herd size, breed types, and region were included in the model as covariates to control for confounding effects associated with these variables. Average age was excluded because it was likely related to the herd replacement rate, which could be affected by BVDV infection status, and thus absorb some of the variability in the response that was actually associated with BVDV. The model was
 |
where Y was the response variable (pregnancy rate, first-service conception rate, calving-conception interval, abortion rate, induction rate, proportion physiological service intervals, culling rate, standard milk per day, logSCC); 
was the intercept; %INH was either the inhibition percentage or the BVDV antibody class (1 to 6), herdsize was the number of recorded calvings in the season 2001–2002; HF, JF and CR were the proportions of breed type in each herd (HF = Holstein-Friesian, JF = Jersey-Friesian, CR = crossbreed). Region was a categorical variable for the location of herds (Northland, Bay of Plenty, and Waikato).
Adjusted means (geometric means for SCC) and confidence intervals were computed as least square means (LSM) and compared between each of the 6 groups of BTM antibody against BVDV (%INH) using the method of Hsu and Nelson (1998) to correct for multiple comparisons in GLM.
For response variables measured as a proportion, herds including less than 30 cows were not included in the calculations for 2 reasons: 1) to adjust for the relatively large uncertainty about mean proportions based on a low group size, and 2) to remove herds with potential selection bias if only a small fraction of cows was included in the mean estimate, inferring that the mean may not be representing the true herd average.
Plots of residuals vs. predicted values of all models were examined for consistency with assumptions for regression analysis; that is, normal distribution and homogeneity of variances. Independence of the study herds with respect to the performance outcomes was examined (ratio of residual Pearson 
2/degrees of freedom error >1); models were adjusted for lack of independence using this ratio as a scale parameter (McCullagh and Nelder, 1989).
Partial Budget
The results of information obtained from the current study population were combined with data from other sources to calculate a partial budget estimate for the production loss of an average dairy cow in a herd with BTM antibody against BVDV of >80%INH compared with 80%INH. The result was extrapolated to the entire dairy herd population in New Zealand based on an estimated prevalence of herds with PI animals of 14.6% (Thobokwe et al., 2004).
The calculations included loss because of extended calving-to-conception intervals (+2.35 d), a reduction in milk solids production (–0.074 kg/d for 250 d) adjusted for less pasture use equivalent to the proportion of lost production, an increase in the annual abortion rate (+2.03%), reduced survival to 24 mo, and increased replacement rate of PI calves with unobserved BVDV infection (Voges, 2006). Values of milk solids and livestock at different ages and production status were sourced from various industry Web sites (http://www.agridata.co.nz/calf-bobby.asp; http://www.agridata.co.nz/cow-dairy.asp; http://www.agrida-ta.co.nz/Feed.asp; www.ifcndairy.org/specialstudies/2005/2005-07.pdf). The price of a replacement heifer ($1,200) was discounted by the average value of a cull cow ($324). The value of a calf at 100 kg was discounted by the daily grazing cost for calves ($0.9/d for 120 d), setting other maintenance costs to $0.1/d. The cost of nonproductive pasture days by nonpregnant, dry cows was assumed to be 120 dry-days x $2.4/d x 2.03% extra abortions. The cost of larger induction rates on affected farms was calculated as half the net benefit of calves at 100 kg of BW, assuming that half of all inductions would lead to a calf born dead.
The calculations included a proportion of 1.33% PI calves in an average affected herd (Voges, 2006). This estimate was derived from a cohort of 904 reared heifers from 300 herds, 12 of which were PI. In seasonal mating herds (about 92% of New Zealand herds), the real proportion of PI in an affected herd will often be much larger if susceptible cows are exposed to PI animals during the mating period of approximately 2 mo. The percentage PI in affected herds therefore varied from 1 to 10% in a sensitivity analysis. The calculation of loss among PI animals included shorter survival up to 2 yr of age and a 5.6-times increased culling rate of PI cows entering the lactating herd (Voges, 2006).
Not considered in the loss estimate was additional health cost because of disease in calves and adult cows (Fourichon et al., 2005).
To include uncertainty in the calculation of the partial budget estimate caused by the cumulative variation of all input parameters, a stochastic simulation using Latin Hypercube sampling with 10,000 iterations was performed in @Risk (Palisade Corporation, 2005). Uncertainty about the input variables was specified as standard deviation for normally distributed variables and as minimum, most likely, and maximum values for variables assumed to be pert-distributed. The result was displayed as a cumulative distribution and 5th to 95th percentile of the average annual loss per 251-cow herd. Moreover, standard regression coefficients were computed from a multifactorial regression of the outcome (annual loss per herd) on all input variables to compare the relative impact of these variables on the loss estimate (sensitivity analysis).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
. Half of all herds did not report any abortions and half did not report any inductions (indicated by a median of zero).
|
). Even though the annual rates of abortion and induction were obviously not normally distributed, the model residuals had a normal distribution and their variances were homogeneous across the range of predicted values.
Association Between Production Loss and BTM Antibody Concentration
Least square means (±SE) for each of the 6 BVDV BTM antibody classes adjusted for herd size, region, and breed composition are shown in Table 2. Not all performance indicators were associated with the concentration of BVDV antibody in BTM. For example, the first-service conception rate, the rate of physiological service intervals, the number of services per conception, the annual pregnancy rate, the geometric herd average SCC, the overall culling rate, and the culling rate on account of conception failure were not associated.
|
). This was calculated as the mean for herds with >80%INH minus the mean for all other groups. Abortion rates were 1.63% (SE 0.76%) to 1.85% (SE 0.85%) when the concentration of BVDV antibody in BTM was 60%INH and 3.36% (SE 1.02%) to 4.42% (SE 0.80%) when >60%INH (Table 2). This cut-off point for %INH was therefore chosen to estimate the effect of BVDV on abortion but only considered for herds with BTM antibody >80%INH in the partial budget. The increase in the high BTM antibody group was 2.03% (SE 0.69%). The rate of inductions, which were reported to LIC, was substantially larger (0.44%; SE 0.12%) in highly exposed farms (>80%INH) than in all other farms (0.12%; SE 0.05%). The fact that overall induction rates were generally low was very likely caused by a low reporting rate.
|
80%INH (Figure 2). This was equivalent to 0.074 kg of milk solids per day and 5.8% of total production.
|
translates the adverse effects of a high BVDV antibody concentration in BTM observed in this study to gross economic loss in affected herds (>80%INH vs. 80%INH) and extrapolates this estimate to the level of the dairy industry of New Zealand. For an average herd with 251 cows, the largest annual loss component in affected herds was the decline in milk production (NZ$10,516), followed by increased abortions (NZ$5,935), increased time to conception (NZ$3,369), loss among PI animals (NZ$1,660), and calf loss as a consequence of abortion or induction (NZ$441). This summed to a total annual loss of about NZ$21,921 per year for a herd with 251 cows affected by BVDV, with a 90% credible interval of NZ$17,959 to NZ$29,666 (Figure 3). Given that 14.6% of all herds in the 3 study regions had BVDV antibody concentrations of >80%INH, and if it is assumed that this is similar in all regions of New Zealand, the estimated loss to the dairy industry in New Zealand amounts to approximately NZ$44.5 million every year (credible interval of NZ$32.3 to NZ$65.1 million).
|
|
). This was related to the large uncertainty (1 to 10%) assumed for this variable for reasons mentioned previously. A change in the proportion PI from 1.3 to 2.7% increased the total loss by 7%. In order of relative impact, the proportion PI was followed by the farmgate price for milk solids, the effect of BVDV exposure on milk production, increased abortions, herd average milk production, the cost per pasture-day, and the cost of replacement heifers. All other factors had relatively little impact on the loss estimate.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
may be regarded as typical for the New Zealand dairy industry. Reporting bias may have existed, however, because the data were extracted from the LIC database. The bias may particularly affect abortion and induction rates because farmers are unlikely to observe and report such events accurately. Induction rates in particular were far below expectation. Abortion rates, however, were mainly derived from the presence or absence of dates of subsequent calving, and the mean rate of 2.6% was similar to that in an earlier report (Thobokwe and Heuer, 2004). Although the reported induction rates may be grossly underestimated, it was unlikely that the rate of underreporting correlated with BTM antibody to BVDV; hence, we believe that the observed association between induction rates and BTM antibody to BVDV was valid. Using herd averages for both, the extent of exposure to BVDV and performance parameters combined with estimated economic loss requires that the implied causal associations observed at the herd level were also true at the animal level. The analysis of this study used data from observations at the herd level, thus we cannot prove that the same associations existed at animal level. The same constraint applies to several studies in which the BVDV status and the economic response have also been defined at the herd level (Fredriksen et al., 1998; Valle et al., 2001; Robert et al., 2004). As long as the BVDV status is defined at the herd level, any analysis may be prone to ecologic fallacy; that is, the inference from group- to individual-level associations. The benefit of carrying out analysis at the herd level, however, is that data can be obtained from a subset of the reference population thus allowing inference for the dairy industry at reasonable cost. Because it was the primary objective of this study to estimate industry-wide losses associated with BVDV in dairy herds in New Zealand, we decided to use a herd-level analysis to provide initial estimates of loss associated with BVDV to be refined by animal-level investigations later. We believe that our results were especially credible for those response variables that showed a trend effect of decreasing performance with increasing BTM %INH.
The results clearly showed that the BVDV antibody concentrations of BTM correlated with reproductive and production performance, and these associations have also been shown to exist at animal level by other workers mentioned earlier. Thus, it seems highly unlikely that a strong ecologic fallacy biased the results of this study. The group of herds with >80%INH exhibited poor performance. This was the group that showed a close relationship between BTM antibody and antibody prevalence in calf mobs (i.e., spot-check) with 81% sensitivity and 91% specificity (Thobokwe et al., 2004). The spot-check is a recognized method to determine whether one or more PI animals are likely to be present in the milking herd (Houe and Palfi, 1993; Houe, 1994; Bitsch and Rønsholt, 1995). Nevertheless, misclassification of the BVDV herd status based on a single antibody test in BTM remained a concern. At a relatively low prevalence of actively infected herds, the consequence of a BTM test with the given sensitivity and specificity would be an overestimate of true herd prevalence. This bias, however, was probably small and, combined with a likely underestimate of the economic effects (e.g., attributable to excluding effects of BVDV on clinical disease incidence), the resulting bias on economic consequences of BVDV at population level was probably very small.
The rate of physiological service intervals was analyzed because BVDV is known to cause early embryonic death leading to a return to estrus (McGowan et al., 1993b), but our data did not show an association with BVDV antibody in BTM, possibly because such an effect was too rare to be significant at the population level. The first-service conception rate was expected to decrease with increasing BVDV antibody because in vivo studies have demonstrated fertilization failure to be a primary cause of reduced conception rates (Grahn et al., 1984). We did not observe this effect, possibly because of the small proportion of the herd being infected at the time of fertilization. The annual pregnancy rate is the result of reproductive performance and culling, and it involves repeat breeding. All of these factors contribute to the final reproductive outcome; thus, a BVDV effect was unlikely associated significantly with the overall annual pregnancy rate. Similarly, culling rates were unaffected because only about 40% of the cull cows were removed due to reproductive failure. Similarly, crude culling risk was not increased in BVDV-infected herds in Norway (Valle et al., 2001). Immune suppression leading to decreased neutrophil function in periparturient disorders such as mastitis in cows (Cai et al., 1994) was hypothesized, but not substantiated, by the data. In earlier studies, increasing or persistently high bulk-tank BVDV status has been associated with an increased incidence of clinical mastitis, although no effect on SCC was observed in this or earlier studies (Niskanen et al., 1995; Waage, 2000).
The time from calving to conception increased probably because of fertilization failure and embryo loss (Grahn et al., 1984; McGowan et al., 1993a). We observed increased abortion rates when BVDV antibody concentrations were greater than 60%INH. Other studies also found an increased number of abortions in herds experiencing natural infection with BVDV (Barber et al., 1985; Roeder et al., 1986; Larsson et al., 1994). Moreover, abortions could be induced by experimental infection (Casaro et al., 1971; Done et al., 1980). The association between herd-level BVDV infection status and abortion rates was therefore not surprising. A delay in conception may explain why induction rates were substantially larger in herds with >80%INH of BVDV antibody in BTM.
A sudden reduction in milk yield has been described from individual herd disease outbreaks (Barber et al., 1985; Larsson et al., 1994). Moerman et al. (1994) found a significant reduction in milk yield experienced in cows seroconverting to BVDV compared with herd mates that did not seroconvert. Thus, immune suppression followed by an increase in clinical mastitis (Cai et al., 1994) and feed energy used for immune function (both initiated by BVDV infection) were plausible causes to explain the observed decline in milk production.
Information on mating dates was scantier than other performance data and missing from many large herds. Thus, the estimate of 91.2% physiological service intervals with a standard deviation of 5.7% may be valid for herds with up to 500 cows. The estimated 1.40 (SE 0.24) services per conception was probably an underestimate of the true number of services because more (natural) services will have occurred than are captured in the database. The mean induction rate was likely underestimated, because herd owners might have been reluctant to report the true number of induced cows to LIC. All other performance indicators and their standard deviations seemed to be reasonable estimates for dairy herds in the 3 study regions.
The final loss estimate based on the partial budget approach was probably conservative because, first, it did not include increased incidence of clinical disease such as calf disease, mastitis, or retained placenta (Larsson et al., 1994; Niskanen et al., 1995). In French dairy herds, the inclusion of disease in a partial budget increased the loss estimates by 10 to 16% (Fourichon et al., 2005). The generally smaller disease incidence in New Zealand dairy herds may not affect our estimates greatly, but including disease would certainly increase the estimate of loss. Second, the sensitivity analysis showed that the proportion of PI animals in the herd had a substantial effect on economic loss based on the particular uncertainty about this parameter. This was substantiated by case reports of >10% PI heifers in a herd (Ellison et al., 2005; Thompson, 2005). Third, misclassification may have included a number of noninfected herds in the exposed group (>80%INH) and infected herds in the nonexposed group (0 to 80%INH). This would reduce the true difference between infected and noninfected herds. We therefore believe that the total loss of NZ$44.5 million per 3.48 million dairy cows, equivalent to NZ$13.7 million per 1 million calvings, may be an underestimate. It is certainly in the lower range of loss estimated for the US; that is, US$10 to US$57 million (Houe et al., 1993a,b; Houe, 1999).
The 2 factors with the greatest impact on this estimate (also the factors about which no reliable information exists to date) were the true prevalence of herds with an active infection and the proportion of PI heifers and cows in the lactating herd. A 1% increase in the proportion PI was associated with NZ$1.55-million-greater loss at the population level; thus, valid estimates of the proportion PI in infected herds would have substantial impact on the accuracy of total loss.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Received for publication April 4, 2007. Accepted for publication August 19, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
