An integrated approach to assessing the viability of eradicating BVD in Scottish beef suckler herds

HAL Id: hal-00578409

https://hal.archives-ouvertes.fr/hal-00578409

Submitted on 20 Mar 2011

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

An integrated approach to assessing the viability of eradicating BVD in Scottish beef suckler herds

B.J.J. Mccormick, A.W Stott, F. Brülisauer, B.V. Ahmadi, G.J. Gunn

To cite this version:

B.J.J. Mccormick, A.W Stott, F. Brülisauer, B.V. Ahmadi, G.J. Gunn. An integrated approach to

assessing the viability of eradicating BVD in Scottish beef suckler herds. Veterinary Microbiology,

Elsevier, 2010, 142 (1-2), pp.129. �10.1016/j.vetmic.2009.09.053�. �hal-00578409�

Accepted Manuscript

Title: An integrated approach to assessing the viability of eradicating BVD in Scottish beef suckler herds

Authors: B.J.J. McCormick, A.W Stott, F. Br¨ulisauer, B.V.

Ahmadi, G.J. Gunn

PII: S0378-1135(09)00470-2

DOI: doi:10.1016/j.vetmic.2009.09.053

Reference: VETMIC 4610

To appear in: VETMIC

Please cite this article as: McCormick, B.J.J., Stott, A.W., Br¨ulisauer, F., Ahmadi, B.V., Gunn, G.J., An integrated approach to assessing the viability of eradicating BVD in Scottish beef suckler herds, Veterinary Microbiology (2008), doi:10.1016/j.vetmic.2009.09.053

This is a PDF file of an unedited manuscript that has been accepted for publication.

As a service to our customers we are providing this early version of the manuscript.

The manuscript will undergo copyediting, typesetting, and review of the resulting proof

before it is published in its final form. Please note that during the production process

errors may be discovered which could affect the content, and all legal disclaimers that

apply to the journal pertain.

Accepted Manuscript

An integrated approach to assessing the viability of

1

eradicating BVD in Scottish beef suckler herds

2 3 4 5

B. J. J. McCormick

, A.W Stott

, F. Brülisauer

, B.V. Ahmadi

, G.J. Gunn

* 6

7

* Corresponding Author 8

Tel - ++44 (0)1463 246066 9

Fax - ++44 (0)1463 236579 10

Email. george.gunn@sac.ac.uk 11

12

1

Epidemiology Research Unit, SAC (Scottish Agricultural College), King's Buildings, 13

West Mains Road, Edinburgh, EH9 3JG 14

2

Bio-Economics and Rural Strategy Team, Epidemiology Research Unit, SAC (Scottish 15

Agricultural College), King's Buildings, West Mains Road, Edinburgh, EH9 3JG 16

17

Keywords: bovine viral diarrhoea (BVD), stochastic model, marginal benefits 18

19

Draft for special issue of Veterinary Microbiology from the 7th ESVV Pestivirus 20

Symposium 21

22

Accepted Manuscript

23

Abstract 24

The viability of eradicating bovine viral diarrhoea (BVD) in Scottish suckler herds is 25

dependent on the continued compliance with eradication schemes. At the farm level, the 26

costs of BVD have been identified in previous studies and show a substantial financial 27

imperative to avoid infection. At a regional level the incentives of BVD eradication to 28

individuals are unclear, for example the requirement for vaccination strategies despite 29

achieving disease-free status. Ensuring farmer compliance with an eradication scheme is 30

therefore difficult. Experience of eradicating BVD from beef-dominated areas is limited 31

and theoretical models have tended to focus on the dairy sector. Here we present a 32

stochastic epidemiological model of a typical beef suckler herd to explore the interaction 33

of a farm with a regional pool of replacements, utilising information from a BVD virus 34

seroprevalence survey of Scottish beef suckler herds. Our epidemiological model is then 35

used to assess the relative costs to individuals assuming different regional endemic 36

prevalences, which are used to represent the likelihood of BVD re-introduction. We 37

explore the relative cost of BVD, taken as likelihood and consequence, at an endemic 38

steady state in contrast to previous models that have assumed the introduction or control 39

of BVD in an epidemic state (e.g. a closed and mostly susceptible population). Where 40

endemic, BVD is unlikely to affect all farms evenly and will cost most farmers very little 41

due to herd immunity or self-clearance of the virus. Compliance is likely to be boosted by 42

pump-priming to initiate and complete eradication schemes with cost-sharing.

43

44

Accepted Manuscript

Introduction 45

Bovine viral diarrhoea (BVD) is an endemic livestock disease causing substantial 46

economic losses in dairy and beef cattle herds worldwide (Bennett and Ijpelaar 2005).

47

The biggest losses occur through reduced fertility and the generation of persistently 48

infected (PI) animals that are a source of continued infection and can suffer high 49

mortality from the development of mucosal disease (Brownlie 1985; Roeder and Drew 50

1984). BVD virus (BVDV) causes immunosuppression (Chase et al. 2004), predisposing 51

animals to a wide range of disease and resulting losses will often be attributed to 52

secondary causes. As a consequence, economic as well as animal health impacts of BVD 53

are often underestimated in herds with endemic infection.

54 55

A comparative study of the economic challenge of BVD in several European countries or 56

regions (Gunn et al. 2005), suggests that the UK may be at a relative disadvantage both in 57

terms of the current challenge from BVD and in the costs to individual farmers of 58

unilateral attempts to establish and maintain disease freedom. This highlights the need for 59

a more detailed study of the economics of BVD at regional rather than just at individual 60

farm level. Regional control has tended to focus on areas dominated by the dairy 61

industry. This is not the case in Scotland, which has a much larger beef sector and where 62

there is limited practical experience of eradicating BVDV from beef herds, although 63

some success has been had on the Shetland and Orkney islands (Synge et al. 1999).

64 65

Based on the guidance of the OIE (2008), we use a risk analysis framework to integrate 66

studies from complimentary disciplines to explore the feasibility of a regional eradication

67

Accepted Manuscript

scheme in the context of Scottish beef suckler herds. McInerney (1996) has suggested 68

that the marginal costs of an eradication scheme might increase considerably towards the 69

end as prevalence decreases and it becomes more difficult to find and eliminate the last 70

and most intransigent infected herds. The problems experienced on the Shetland islands 71

during the eradication of BVD (Synge et al. 1999) lend weight to this argument. Marginal 72

costs at the start of an eradication programme may also be higher as maintaining disease- 73

free herds will be more expensive while regional prevalence, and thereby challenge, of 74

BVDV remains high (Gunn et al. 2005). At both extremes of the scheme farms are, 75

therefore, less likely to either embark on or comply with the protocol because the balance 76

of ‘risks’ is less favourable than mid-way through the programme. Identification of the 77

relative cost, or their inverse benefits, to individuals and to the public (i.e. a wider 78

farming community) may enable appropriate ‘beneficiary pays’ cost sharing to pump- 79

prime an eradication scheme at its start and end. However, estimates of the potential 80

extent of this effect are needed to inform policy.

81 82

We explore the financial incentives to control BVD, as an individual farm or a region and 83

the identification of the ‘beneficiary’ who ought to pay based on the level of endemic 84

disease. Endemic disease represents both the status quo, that which is accepted as the 85

norm, and then by reducing the prevalence it represents stages of an eradication scheme 86

which aims to reduce prevalence of disease to zero. For this purpose, we present a 87

stochastic model of a typical Scottish beef suckler herd that interacts with a regional pool 88

of replacement animals of known BVDV prevalence. Previous models (Gunn et al. 2004;

89

Innocent et al. 1997; Viet et al. 2004) have dealt with the incentives to control BVDV on

90

Accepted Manuscript

a single closed farm and have explored the consequences of introducing BVDV into a 91

wholly susceptible farm. Here we present a model of a farm that introduces animals from 92

an external source to explore the longer term consequences of different likelihoods of re- 93

introducing BVDV. Our model also explores the endemic state of BVDV, at which point 94

there is an approximately stable prevalence of infected herds even though for any given 95

farm the incidence of BVDV is very low.

96 97

Materials and Methods 98

Data 99

BVD seroprevalence in Scotland 100

The starting point for an eradication programme is to identify the prevalence level in 101

order to assess the scale of the challenge and to establish a benchmark from which to 102

estimate the changing marginal benefits of eradication postulated above. This study is 103

based on the results of a survey estimating the national prevalence of active BVDV herd 104

infection in Scottish beef suckler herds (Brülisauer et al. 2009).

105 106

A survey comprising 301 beef suckler herds, stratified by geographic region and herd 107

size, was carried out in 2006-2007. Blood samples were taken from a randomised sample 108

of between seven and ten youngstock (6 – 16 months old) from each management group 109

within the enterprise. Samples were tested by BVDV antibody ELISA and information 110

gathered on farm characteristics and herd management factors using a standardised

111

Accepted Manuscript

113

One third of beef suckler herds showed evidence of recent BVDV exposure, with at least 114

one animal testing BVDV seropositive. Sixteen percent (95CI: 11.6, 19.7%) of herds had 115

youngstock with a seroprevalence of more than 90%. This group therefore showed 116

evidence for recent active BVDV infection and presently or in the past harbouring a PI 117

animal.

118 119

Modelling 120

Epidemiological modelling 121

Following Gunn et al. (2004), we modelled a typical Scottish 100 cow beef suckler herd, 122

using a state transition approach for the BVDV dynamics in an age stratified population.

123

Our stochastic simulation model was designed to account for the small finite population 124

of beef cattle and featured a daily, rather than annual, time step to allow greater emphasis 125

on the timing of (re-)introduction of BVDV via the purchase of animals. These features 126

allowed us to better explore the importance of variability in the timing and likelihood of 127

introduction of infectious animals from off-farm relative to rebreeding and other 128

significant management activities on-farm. We modelled BVDV dynamics on a single 129

typical farm using no BVDV control measures, which interacted with a regional pool of 130

herds that were the source of replacement animals. The proportion of animals in each of 131

the mutually exclusive disease states (susceptible, S; transiently and persistently 132

infected/infectious, TI and PI respectively; and immune animals, R) on the farm was 133

dynamic based on the assumptions in Table 1.

134

Accepted Manuscript

135

[Table 1]

136 137

Regional pool properties 138

The regional pool was designed as a collection of replacement animals and modelled as a 139

separate set of disease-state classes. Both the number of animals and the prevalence of 140

BVDV in the regional pool were assumed to be static during a given model scenario. The 141

rationale was to mimic the risk of BVDV infection/re-infection to which an individual 142

farm is exposed when buying replacements from a large pool of other farms (the region) 143

assuming an endemic (i.e. steady) prevalence of BVDV equal to, or below, the current 144

level. The regional herd in this model represented any large collection of available 145

animals from a local cooperative to a national herd that can be assumed to have relatively 146

stable, endemic, BVDV prevalence. The assumed stability of disease is supported by the 147

apparent consistency of BVDV prevalence across different countries (Houe 1999), 148

suggesting that in spite of the management, spatial and temporal heterogeneity, BVDV is 149

relatively stable at a regional level.

150 151

We modelled different scenarios, each assuming a different prevalence of PIs in the 152

regional pool, thus the probability of acquiring a PI was a function of the regional 153

prevalence of PIs. The rest of the animals in the regional pool were assumed to be 154

susceptible. This assumption served two functions; first it represents the worst-case 155

scenario of buying either highly infectious PI animals, which are considered the primary 156

source of BVDV on farms (Houe 1999), or susceptible animals that can then become

157

Accepted Manuscript

infected and do not contribute to the herd immunity. The presence of antibodies are not 158

necessarily an indicator of immunity as they can arise from vaccination and they are not a 159

reliable indication of the disease status of offspring which can be infected as foetuses 160

before the cow recovers and develops an immune response. Second, complications 161

associated with the changing risks of buying-in susceptible, immune, transiently infected 162

or PI animals as an eradication programme progresses are eliminated. Although this level 163

of complexity is involved in practice, we felt that the key driver would be the likelihood 164

of purchasing PIs in an endemic situation.

165 166

Farm herd properties 167

Individual cows were modelled to utilise the higher temporal resolution of this model 168

compared to Gunn et al. (2004), which modelled age cohorts. Cohorts were still retained 169

for heifers and calves, but the individual level was deemed more appropriate for breeding 170

cows since this allowed a more detailed exploration of the specific timing of transient 171

infection during pregnancy and thus the probability of different outcomes of foetal 172

infection (see the supplementary information for further details).

173 174

We assume that the population is managed to maintain an approximately constant number 175

of breeding cows (100) (Gunn et al. 2004), and stochastic fluctuations in the population 176

size simulate additional births, deaths or management decisions to replace individual 177

animals. We have assumed that deaths not directly attributable to BVD are accounted for 178

in proportion to the population replaced during management activities (i.e. to maintain a 179

constant population size). These additional deaths may include both mortality from

180

Accepted Manuscript

unrelated causes and secondary infection resulting from BVDV induced 181

immunosuppression (Brownlie et al. 1987).

182 183

Additional mortality resulting directly from persistent infection with BVDV is explicitly 184

modelled. We have assumed an exponential probability of death capturing both those that 185

are born weak and thereby predisposed to secondary pathogens and the mortality 186

resulting from mucosal disease in older PIs (Duffell and Harkness 1985). Mortality was 187

calculated according to the age of PIs reported by Rüfenacht et al. (2000) using the best 188

fitting exponential distribution (mean age of 342.3 days). This approximation produces 189

similar results to those observed by Duffell and Harkeness (1985) with 65% of PIs dying 190

in their first year and 20% surviving to 18 months (Houe 1992). It is worth commenting 191

that Rüfenacht et al. (2000) reported the age of animals observed rather than their age of 192

death, therefore if incorrect, the longevity of PI animals is underestimated (see the 193

supplementary information).

194 195

In our model, cows were put to the bull for a fixed period of 85 days approximately 196

following the upper limit of the recommend window for maximal productivity given by 197

Caldow et al. (2005), and the probability of becoming pregnant was uniform within this 198

period.

199 200

A proportion, 15%, of adult cows was replaced each year assuming an average herd life 201

of seven years (SAC 2007). Replacements were home-bred in the first instance. The 202

availability of home-bred replacements was dependent on the number of calves (and

203

Accepted Manuscript

thereby heifers) retained at the point when calves were sold and replacements potentially 204

bought. If there were insufficient calves to meet the needs of the replacements, as resulted 205

from stochastic fluctuations around the mean population size at the point of weaning, an 206

appropriate number of animals were bought from the regional herd. In accordance with 207

the prevalence survey, it was assumed that 80% of replacements were in-calf as this 208

strategy (and cows with calf-at-foot) represents a major risk of (re-)introducing BVD 209

onto the farm and thus are the most conservative worst-case estimates of marginal 210

benefits. Replacement animals were assumed to be heifers, because it was deemed a more 211

common strategy to replace older animals with younger ones where possible since 212

younger animals would be worth more over time.

213 214

BVDV transmission 215

Transmission was based on Viet et al. (2004) who modelled a dairy herd using a 216

proportionate mixing term to describe the probability of encountering an infectious 217

animal as a proportion of the herd. We assumed that in a suckler herd all animals ran 218

together (Gunn et al. 2004) and therefore the probability of interacting with PIs from 219

different age cohorts were equally likely. The daily mean probability of BVDV 220

transmission was the proportion of PIs in the age cohort multiplied by the force of 221

infection (i.e. the rate at which susceptible animals get infected), 0.5 (Viet et al. 2004).

222

Similarly, virus transmission by transiently infected animals was assumed to be the 223

proportion of TI animals and a force of infection of 0.03 (Viet et al. 2004).

224

225

Accepted Manuscript

Infection of dams during pregnancy resulted in foetal consequences dependent on the 226

month of gestation. The probability of different outcomes (generation of a PI calf, 227

abortion, the birth of a weak calf with congenital defects or no deleterious effect) were 228

based on calculations made by Gunn et al. (1998) using data from Houe (1996) (see the 229

supplementary information).

230 231

As used by Gunn et al. (2004), the initial conditions of the model was a population with a 232

single PI cow, ten TI cows and eight TI heifers along with a further 89 susceptible cows 233

and 18 susceptible heifers.

234 235

Models were run for different regional PI prevalence scenarios (5-0% in 0.05 decrements) 236

in order to cover a range of possible stages of BVDV prevalence from an endemic state 237

and through to eradication. The simulation was run for thirty years, which allowed the 238

first 15 years to be disregarded as the burn-in period and reach a representative steady 239

state (i.e. endemic BVDV). Each scenario was repeated 1000 times to account for 240

stochastic variation.

241 242

Economic modelling 243

The cost of disease health and labour consequences were derived from Stott and Gunn 244

(2008). The cost of BVD was partitioned into five components, namely: (1) the cost of an 245

abortion from a transiently infected cow, (2) the cost of a weak calf from a transiently 246

infected cow, (3) the cost of a PI calf (from a TI or PI cow), (4) the cost of transient 247

infection in calves and (5) the cost of a PI cow. The cost of an embryonic re-absorption

248

Accepted Manuscript

return to service, which constituted the sixth cost in the Stott and Gunn (2008) model was 249

not calculated in this model. This was because the daily time step used in this model 250

meant it was possible to account for the delay in rebreeding and hence in calving via the 251

costs attributed to the calf rather than the cow. Therefore, including the costs of delayed 252

conception would then constitute double counting. More detailed exploration of the 253

interaction between BVD and infertility in suckler cows is undertaken by Varo Barbudo 254

et al. (2008). We also excluded the fixed cost of immunosuppression in calves included 255

by Stott and Gunn (2008) on the grounds that there is insufficient information to establish 256

how this might vary with regional prevalence of BVDV. Our objective was to establish 257

how the marginal cost (rather than absolute cost) of BVD changes as regional prevalence 258

alters at different stages of an eradication programme. The costs of BVD in each run of 259

the simulation model were summarised as the sum of all costs given above in years 15 to 260

30 of the simulated epidemic, discounted in the normal way (Boehlje and Eidman 1984) 261

so that the total discounted costs (TDC) were expressed in present value terms. The 262

discount factor used was 0.05 as in previous economic impact assessments of BVD in 263

Scottish beef suckler herds (Gunn et al. 2004; Stott and Gunn 2008). As cashflows were 264

summed across years 15 to 30 of a simulated epidemic to represent the endemic disease 265

situation and then discounted to present values for comparative purposes they were not 266

comparable with published costs and benefits that are concerned with the more 267

immediate costs of BVD in the early phases of an epidemic. Furthermore, our interest 268

was in the relative incentive to deal with the disease under different regional 269

epidemiological circumstances. Absolute costs are, in any case, dependent on price

270

Accepted Manuscript

assumptions made, which are likely to date quickly. For these reasons, we expressed all 271

our financial results relative to the lowest TDC obtained.

272 273

For simplicity, any costs were incurred when the animal/event occurred. Whilst the cost 274

of replacing a PI cow that is suffering mucosal disease (i.e. the loss of the cow, the cost of 275

replacing an animal and additional labour and veterinary costs) may occur some years 276

after the PI cow comes onto the farm, in this model the charge was noted at the point of 277

occurrence to avoid double counting BVD related costs.

278 279

Results 280

The initial conditions of the model had a dramatic impact on the first 5 years of each 281

iteration, however, these effects were negligible after 10 years. This outcome was in 282

accordance with the observation of previous simulation models that observed self- 283

clearance of BVDV after 7 years (Humphry et al. 2005). We only, therefore, consider the 284

last 15 years of the simulation (i.e. the first 15 years are discarded) such that we only 285

explore the endemic (steady) state of BVDV infection in the modelled herd.

286 287

The median incidence of PIs, that is the average number of PIs occurring on the farm per 288

year, was a linearly function of the regional prevalence of PIs (p <0.001) (Figure 1).

289 290

[Figure 1]

291

292

Accepted Manuscript

The median proportion of simulation time (years 15 to 30) for which BVDV was active 293

(i.e. a non-zero count of PIs or TIs in the herd) was linearly related to the regional 294

prevalence of PIs (p <0.001). There was a strong correlation between the proportion of 295

time BVDV was active and the incidence of PIs in the herd (r 0.972), although the 296

correlation decreased as the regional prevalence of PIs increased and the variability of 297

time active BVDV increased, ranging from an inter-quartile range of 0.20 of simulation 298

time with no PI re-introductions, to 0.58 with a PI prevalence of 5%.

299 300

[Figure 2]

301 302

Assuming that each iteration of the simulation could be treated as a uniquely realised 303

farm, comparison was made between the regional PI prevalence scenarios and the result 304

of the seroprevalence survey for validation of the model results (Figure 2). The 305

seroprevalence survey identified three herd cohorts: seronegative herds, herds with low 306

seroprevalence (1-8 animals seropositive and a median within herd prevalence of 0.3) and 307

high seroprevalence (9 or 10 animals seropositive and a median within herd prevalence of 308

0.9). Each of 1000 iterations of each of the regional prevalence scenarios were 309

categorised into similar cohorts based on a single spot sample of youngstock taken in 310

year 15, effectively replicating the seroprevalence spot sample with the model data. The 311

simulated spot sample identified the proportion of youngstock (calves) in the immune (R) 312

class just as the serology identified the immune status of real livestock. The distribution 313

of herd seroprevalence for each scenario (columns 2-12 on Figure 2) was then compared 314

to the distribution of farm seroprevalence results from the survey (first column on Figure

315

Accepted Manuscript

2) and each of the model scenarios was significantly different (Kolmogorov-Smirnov test 316

at 5%). The best-fitting (i.e. least significant difference, p = 0.0396) regional PI 317

prevalence was 2.5%. This is within the range of previous estimates for PI prevalence 318

from other studies (in other countries, e.g. (Rüfenacht et al. 2000)).

319 320

[Figure 3]

321 322

The regional prevalence of PIs, rather than the herd seroprevalence (as defined by the 323

simulated seroprevalence sampling), dominated the relative TDC due to BVDV (Figure 324

3). For clarity, the different regional PI prevalence scenarios were grouped into three 325

broad groups of high, medium and low prevalence, within which were the simulated herd 326

seroprevalence spot samples. The TDC for each region and herd grouping was compared 327

to the mean of the lowest TDC (low regional PI prevalence and low herd seroprevalence).

328

The TDC in regions of high PI prevalence is as much as 7.3 times the value in regions 329

with low PI prevalence. This result holds irrespective of the herd seroprevalence 330

category. The variance of the TDC is a consequence of the variable epidemiological 331

outcomes, and reflects the processes of infection, self-clearance and re-infection, or a 332

failure to self-clear BVDV. Re-infection into susceptible herds dramatically increases the 333

cost of BVDV in the low prevalence scenarios, increasing the costs as much as 7 times 334

the mean for less than a quarter of holdings whilst the median cost is zero (most farms 335

have self-cleared BVDV and subsequently not re-introduced PIs). A higher proportion 336

(>25%) of farm iterations did not achieve a zero cost in the high regional prevalence

337

Accepted Manuscript

scenarios. The range of TDC is highest in the high PI prevalence scenarios, with an inter- 338

quartile range of as much as 2 to 10 times the mean TDC of the low prevalence scenarios.

339 340

Discussion 341

BVDV infection has been shown to be costly to individual farmers, though the financial 342

argument at the farm level in mainland Scotland is apparently inadequate to persuade 343

farmers (Gunn et al. 2005) to embark upon systematic control as has been adopted in 344

other regions (Lindberg et al. 2006).

345 346

The results of simulating BVDV dynamics using the initial conditions was highly 347

variable and produced results that contrasted markedly with results of the endemic state 348

of BVDV in subsequent years of the simulation. This observation challenges the reliance 349

on previous models (Cherry et al. 1998; Gunn et al. 2004; Innocent et al. 1997; Viet et al.

350

2004) that have reported findings for which interpretation should be restricted to 351

‘epidemic’ BVDV, that is to say, the spread of BVDV in a wholly susceptible population.

352

In Scotland, BVDV is considered an endemic disease and modeling farms that interact 353

with one another and thereby retain a variable degree of herd immunity, is more realistic 354

than a susceptible and closed herd. The nature of stochastic models may also yield 355

variable results at the outset of a simulation that do not adequately represent patterns that 356

form later in the model as the simulation begins to equilibrate. This is a danger if a 357

sufficient burn-in period is not used and also if it is not discarded from subsequent 358

analysis.

359

360

Accepted Manuscript

Modelling with the possibility of BVDV re-introduction via replacements is something 361

novel in BVDV literature, which has tended to focus on closed herds. The dynamics 362

presented here do not capture the trading patterns of any given farm, and are in fact 363

rudimentary. They do, however, allow for the potential re-introduction of BVDV through 364

interaction with an external source of infection. In the case that the trading, here 365

described as purchasing replacements, is deemed inadequate, the principal of interacting 366

with a wider environment may be interpreted as any introduction of PI animals. Future 367

work will undoubtedly focus on the interaction of farms, given that for individuals there 368

is already a known financial imperative to act. Examination of the role of neighbours, and 369

therefore the necessity to cooperate is now required.

370 371

The regional prevalence of PIs dominated the financial consequence of BVDV in this 372

simulation. This was a surprising result given the high cost to a single farm (Gunn et al.

373

2004) and the potential for BVDV to linger in herds for protracted periods of time 374

(Brownlie et al. 1987). However, previous studies both observational (Stahl et al. 2008) 375

and simulated (Humphry et al. 2005) have identified the natural extinction of BVDV.

376

This process, dubbed ‘self-clearance’ (Lindberg and Houe 2005) is likely to reduce the 377

prevalence of BVDV in a closed system, however the re-introduction of infection will 378

either maintain a low level of virus circulating or produce repeated epidemic outbreaks.

379

For any given farm the classification derived from the spot test would change over time, 380

thus the prevalence for the population of farms would be consistent and for the individual 381

farms would fluctuate. At low regional prevalence of PIs the chance of re-introducing 382

infection is low (or zero) such that the cost of BVDV to the region is negligible as farms

383

Accepted Manuscript

successively self-clear. At higher regional prevalence the regional cost is much higher 384

because farms are more prone to re-introducing BVDV.

385 386

Given the regional cost of BVDV there is a strong financial incentive for regions, over 387

and above individual holdings, to cooperate to eradicate BVDV. This is an interesting 388

finding since for individuals there may be high disease cost, but then the vast majority 389

(almost 70% in Scottish Beef suckler herds) are BVDV-free and then have no cost. The 390

increase in the variability of the TDC for the high regional PI prevalence suggests that the 391

cost of BVDV across the region became more widely distributed across herds, as more 392

herds showed a sustained cost of endemic BVDV over the 15 years of the simulation and 393

fewer herds sustained a disease-free status. Wholly susceptible farms are, however, at risk 394

of epidemic BVDV, which has been shown to incur high costs (Gunn et al. 2004). This 395

dramatic, but brief cost was reflected by the very high TDC for a small minority of farms 396

(<5%) even in the low regional prevalence scenarios compared to highly skewed 397

distribution of zero cost (~95%), which skewed the mean cost for the whole region to a 398

non-zero amount.

399 400

If BVDV is therefore treated as a regional cost, irrespective of any given individual 401

herd’s disease status, then encouraging participation and compliance with an eradication 402

scheme may require pump-priming. With two thirds of farms not currently suffering costs 403

of disease ensuring their active involvement would likely necessitate regional investment 404

(where a region is deliberately left undefined). When BVDV is reduced to low levels 405

modelled here (~0.5% of the population as PIs) the issue is more to do with compliance,

406

Accepted Manuscript

as the proportion of farms with active infection drops to less than 5%. The precise 407

balance of cost-sharing may depend on the scale of the eradication and the mixture of 408

cattle enterprises. Cost-sharing would also require scaling as an eradication scheme 409

progresses since unlike the models presented here, the prevalence of PIs in the regional 410

pool would be a function of the eradication scheme rather than a static risk of BVDV re- 411

introduction.

412 413

Conclusions 414

We have presented a model of endemic BVDV that has explored the increased 415

persistence of this disease on a farm that interacts with a region. Within any given region, 416

for many farms BVDV may not incur a cost given that there is no active transmission of 417

the virus, however the consequence of re-introduction of BVDV to a fully susceptible 418

herd is epidemic BVDV infection with associated increased costs.

419 420

The disease status of any single herd may fluctuate over time, but when iterations were 421

aggregated based on the prevalence scenario, the risk of (re-)introduction of BVDV 422

appears relatively stable assessed across many farms with a similar regional prevalence 423

and has a linear function with the regional prevalence of PIs.

424 425

BVDV can be considered a public issue beyond the financial incentives for individuals to 426

manage the disease. At a regional level, BVD costs all individual farmers (the mean cost 427

to farmers in the region) in higher proportions than the seroprevalence on any given 428

holding. This is likely to result from the dynamic nature of the disease and the balance of

429

Accepted Manuscript

risks of preserving herd immunity, suffering disease losses and self-clearance of the 430

disease (on top of active BVDV control).

431 432

There is a pressing need for greater data to parameterise and structure models for the beef 433

suckler cow industry as most current information is heavily biased towards dairy systems.

434

The relative longevity of animals and the increased mixing both lend themselves to the 435

persistence of BVDV. This is balanced by the dynamics of trading patterns and seasonal 436

calving rather than year-round calving, which may support eradication schemes as 437

potential PIs may be kept apart from pregnant cows at risk of foetal infection.

438 439

Acknowledgements 440

This work was supported by the Scottish Executive Environment and Rural Affairs 441

Department.

442 443

Conflict of interest statement 444

None 445

446

References 447

448

Accepted Manuscript

Bennett,R. and J.Ijpelaar. 2005. "Updated estimates of the costs associated with thirty 449

four endemic livestock diseases in Great Britain: A note." Journal of Agricultural 450

Economics. 56:135-144.

451

Boehlje,M.D. and V.R.Eidman. 1984. Farm Management. Wiley. New York.

452

Brownlie. 1985. "Clinical aspects of the bovine virus diarrhoea/ mucosal disease complex 453

in cattle." In Practice.195-202.

454

Brownlie,J., M.C.Clarke, C.J.Howard, and D.H.Pocock. 1987. "Pathogenesis and 455

Epidemiology of Bovine Virus Diarrhea Virus- Infection of Cattle." Annales De 456

Recherches Veterinaires. 18:157-166.

457

Brülisauer,F., F.I.Lewis, A.G.Ganser, I.J.McKendrick, and G.J.Gunn. 2009. "Prevalence 458

of bovine viral diarrhoea in Scottish beef suckler herds." in submission.

459

Caldow,G., B.Lowman, and Riddell I. 2005. "Veterinary Intervention in the Reproductive 460

Management of Beef Cow Herds." In Practice. 27:406-411.

461

Chase,C.C.L., G.Elmowalid, and A.A.A.Yousif. 2004. "The immune response to bovine 462

viral diarrhea virus: a constantly changing picture." Veterinary Clinics of North 463

America-Food Animal Practice. 20:95-+.

464

Cherry,B.R., M.J.Reeves, and G.Smith. 1998. "Evaluation of bovine viral diarrhea virus 465

control using a mathematical model of infection dynamics." Preventive Veterinary 466

Medicine. 33:91-108.

467

Accepted Manuscript

Duffell,S.J. and J.W.Harkness. 1985. "Bovine Virus Diarrhea-Mucosal Disease Infection 468

in Cattle." Veterinary Record. 117:240-245.

469

Gunn,G.J., H.W.Saatkamp, R.W.Humphry, and A.W.Stott. 2005. "Assessing economic 470

and social pressure for the control of bovine viral diarrhoea virus." Preventive 471

Veterinary Medicine. 72:149-162.

472

Gunn,G.J., A.W.Stott, and R.W.Humphry. 2004. "Modelling and costing BVD outbreaks 473

in beef herds." Veterinary Journal. 167:143-149.

474

Gunn,G.J., A.W.Stott, and S.A.Scanlan. 1998. "Estimating the losses associated with 475

bovine viral diarrhoea (BVD) within the Scottish cow-calf herd." Sydney.1015- 476

1017.

477

Houe,H. 1992. "Age Distribution of Animals Persistently Infected With Bovine Virus 478

Diarrhea Virus in 22 Danish Dairy Herds." Canadian Journal of Veterinary 479

Research-Revue Canadienne De Recherche VeterinaireAlso part of Houe's thesis 480

1996. 56:194-198.

481

Houe,H. Bovinr virus diarrhoea virus (BVDV) epidemiological studies of the infection 482

among cattle in Denmark and USA. 1996. The Royal Veterinary Agricultural 483

University Copenhagen.

484

Ref Type: Thesis/Dissertation 485

Houe,H. 1999. "Epidemiological features and economical importance of bovine virus 486

diarrhoea virus (BVDV) infections." Veterinary Microbiology. 64:89-107.

487

Accepted Manuscript

Humphry,R.W., A.W.Stott, and G.J.Gunn. 2005. "Modelling BVD at herd level 488

compared with individual animal level." Preventive Veterinary Medicine. 72:169- 489

175.

490

Innocent,G., I.Morrison, J.Brownlie, and G.Gettinby. 1997. "A computer simulation of 491

the transmission dynamics and the effects of duration of immunity and survival of 492

persistently infected animals on the spread of bovine viral diarrhoea virus in dairy 493

cattle." Epidemiology and Infection. 119:91-100.

494

Lindberg,A., J.Brownlie, G.J.Gunn, H.Houe, V.Moennig, H.W.Saatkamp, T.Sandvik, and 495

P.S.Valle. 2006. "The control of bovine viral diarrhoea virus in Europe: today and 496

in the future." Revue Scientifique et Technique-Office International des 497

Epizooties. 25:961-979.

498

Lindberg,A. and H.Houe. 2005. "Characteristics in the epidemiology of bovine viral 499

diarrhea virus (BVDV) of relevance to control." Preventive Veterinary Medicine.

500

72:55-73.

501

McInerney,J. 1996. "Old economics for new problems - Livestock disease: Presidential 502

address." Journal of Agricultural Economics. 47:295-314.

503

OIE. 2008. Handbook on Import Risk Analysis for Animals and Animal Products: Volume 504

1, Introduction and Qualitative Risk Analysis. OIE (World Organisation for 505

Animal Health). Paris, France.

506

Roeder,P.L. and T.W.Drew. 1984. "Mucosal Disease of Cattle - a Late Sequel to Fetal 507

Infection." Veterinary Record. 114:309-313.

508

Accepted Manuscript

Rüfenacht,J., P.Schaller, M.Strasser, E.Peterhans, and L.Audigé. 2000. "Prevalence of 509

cattle infected with bovine viral diarrhoea virus in Switzerland." The Veterinary 510

Record. 147:413-417.

511

SAC. 2007. Farm Management Handbook 28th Edition. SAC. Edinburgh.

512

Stahl,K., A.Lindberg, H.Rivera, C.Ortiz, and J.Moreno-Lopez. 2008. "Self-clearance 513

from BVDV infections - A frequent finding in dairy herds in an endemically 514

infected region in Peru." Preventive Veterinary Medicine. 83:285-296.

515

Stott,A.W. and G.J.Gunn. 2008. "Use of a benefit function to assess the relative 516

investment potential of alternative farm animal disease prevention strategies."

517

Preventive Veterinary Medicine. 84:179-193.

518

Synge,B.A., A.M.Clark, J.A.E.Moar, J.T.Nicolson, P.F.Nettleton, and J.A.Herring. 1999.

519

"The control of bovine virus diarrhoea virus in Shetland." Veterinary 520

Microbiology. 64:223-229.

521

Varo Barbudo,A., G.J.Gunn, and A.W.Stott. 2008. "Combining models to examine the 522

financial impact of infertility caused by bovine viral diarrhoea (BVD) in Scottish 523

beef suckler herds." Journal of Agricultural Science. 146:621-632.

524

Viet,A.F., C.Fourichon, H.Seegers, C.Jacob, and C.Guihenneuc-Jouyaux. 2004. "A 525

model of the spread of the bovine viral-diarrhoea virus within a dairy herd."

526

Preventive Veterinary Medicine. 63:211-236.

527

528

529

Accepted Manuscript

530

531

532

Accepted Manuscript

Table 1: Parameter names and values used in the stochastic model. Where a distribution has been

532

used, the values given relate to the distribution parameters.

533

534

Figure 1:The annual incidence of PIs in the farm in the endemic state (the last 15 years of the

535

simulation), showing the median, the 25

and 75

for each scenario with varying regional prevalence

536

of PI animals.

537 538

Figure 2:Comparison of BVDV seroprevalence in the youngstock drawn from Scottish beef suckler

539

herds and modelled regional PI prevalence scenarios during the endemic state. Horizontal lines show

540

the statistical medians of dividing the three cohorts (69% seronegative; 15% 0-60% seropositive;

541

16% 90-100% seropositive) correcting for imperfect diagnostic test interpretation.

542 543

Figure 3: Summary statistics of the distribution of total discounted costs (TDC) due to BVD in years

544

15 to 30. TDCs for each grouped scenario are all relative to the lowest TDC scenario (low herd

545

Parameter Distribution Value Reference

Number of breeding cows (N = S+TI+R+PI)

constant 100 Farm target number of

breeding cows (T)

constant 100 Gunn et al. 2004

Farm target proportion of population changed (C)

constant 0.15 assuming an animal is kept for an average of 7 years

SAC 2007 Births from TI – normal

Births from TI – aborted Births from TI – weakling Births from TI – PI

See supplementary information

Gunn et al. 1998;

Houe 1996

Births from PI constant 1 PI calf/ PI cow/ year Brownlie et al.

1987

Deaths PI calf/heifer exponential mean 342.3 days Rüfenacht et al.

2000 Transmission: PI cow Bernoulli P(success) 0.5

Transmission: PI calf/

heifer

Binomial P(success) 0.03*PI/N,

# trials S Transmission: TI cow Bernoulli P(success) 0.5 Transmission: by PI calf/

heifer

Binomial P(success) 0.03*PI/N,

# trials S

Viet et al. 2004

Recovery of TI cow Bernoulli P(success) 0.091 Recovery: by PI calf/

heifer

Binomial P(success) 0.091,

# trials TI

Cherry et al.

1998 Probability of purchasing

an in-calf animal

constant 0.8 Brülisauer et al.

2009)

Accepted Manuscript

grouped into high (4-5% PIs), medium (2-4% PIs) and low (0-2% PIs), and herd seroprevalence was

547

based on a single spot sample of youngstock, with seroprevalence cohorts defined by the BVDV

548

survey (low, 0 seropositive youngstock; medium, 10-80% seropositive; high, 90-100% seropositive).

549

550

551

552

Accepted Manuscript

Accepted Manuscript

Accepted Manuscript