Preventive Veterinary Medicine 117 (2014) 242–250
Contents lists available at ScienceDirect
Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed
The evolution of the prevalence of classical scrapie in sheep in Great Britain using surveillance data between 2005 and 2012 Mark Arnold a,∗ , Angel Ortiz-Pelaez b a
Animal Health and Veterinary Laboratories Agency (AHVLA), The Elms, College Road, Sutton Bonington, Loughborough, LE12 5RB, UK Epidemiology, Surveillance and Risk Group, Animal Health and Veterinary Laboratories Agency, New Haw, Addlestone, Surrey, KT15 3NB, UK b
a r t i c l e
i n f o
Article history: Received 7 October 2013 Received in revised form 11 July 2014 Accepted 31 July 2014 Keywords: Scrapie Back-calculation methods Prevalence estimation Maximum likelihood methods
a b s t r a c t After the decline of the Bovine Spongiform Encephalopathy (BSE) epidemic in Great Britain (GB), scrapie remains the most prevalent animal Transmissible Spongiform Encephalopathy (TSE) present in GB. A number of control measures have been implemented for classical scrapie, and since 2005 there has been a large reduction in the number of observed cases. The objective of this study is to estimate two measures of disease frequency using up to date surveillance data collected during and after the implementation of different control measures established since 2004, and breeding for resistance schemes that ran from 2001 until 2009. This would enable an assessment of the effectiveness of both the breeding for resistance programme and the compulsory eradication measures in reducing the prevalence of scrapie in GB. Evaluation of the sensitivity of the rapid post-mortem test for scrapie indicated that it detected scrapie in the last 25% of the incubation period. A back-calculation model was developed to estimate the prevalence of infection at animal and flock-level. The results of the model indicated a mean drop of infection prevalence of 31% each year, leading to a 90% drop in infection prevalence between 2005, with an estimate of 5737 infected sheep in GB in 2012. The risks of classical scrapie infection in animals with genotypes of National Scrapie Plan Types I–IV (all other genotypes), relative to Type V (all genotypes containing V136 R154 Q171 and not A136 R154 R171 ), were estimated to be: 0, 0.0008, 0.07, and 0.21 respectively. The model estimated a very low rate of reporting of clinical suspects and a large decline from 2007 of the probability of a sheep being reported as a clinical suspect. The model also estimated that the expected number of sheep holdings with classical scrapie in 2012 was 215 (95% confidence interval: 33–437), out of a total of approximately 72,000 sheep holdings in GB. Model estimates indicate that the prevalence in 2012 has dropped to 10% of that in 2005, showing the effectiveness of the control measures. It also shows a bias in the destination of infected animals, with the majority of infected animals being detected in the fallen stock surveillance stream, and an extremely low proportion of animals detected as clinical suspects; this is very important in terms of the design of surveillance schemes for classical scrapie. Crown Copyright © 2014 Published by Elsevier B.V. All rights reserved.
1. Background ∗ Corresponding author. Tel.: +44 01509 678340; fax: +44 01509 674805. E-mail address: [email protected] (M. Arnold).
After the decline of the Bovine Spongiform Encephalopathy (BSE) epidemic in Great Britain (GB), scrapie remains the most prevalent animal Transmissible Spongiform
http://dx.doi.org/10.1016/j.prevetmed.2014.07.015 0167-5877/Crown Copyright © 2014 Published by Elsevier B.V. All rights reserved.
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
Encephalopathy (TSE) present in the country. The number of clinical scrapie suspects notified by farmers has drastically declined in the last 6 years. Since the disease became a notifiable disease in 1993, 2010 and 2012 have been the only 2 years where no clinical cases of scrapie were confirmed in GB and only three clinical index cases of classical scrapie were confirmed in GB in 2009, and the same number in 2011. The number of clinical cases and the apparent prevalence in the different surveillance streams have shown a continuous decline of classical scrapie over the last 6 years, although no statistically significant differences have been detected year by year (Ortiz-Pelaez et al., 2012). Although surveillance data provide a trend on the evolution of the incidence of the disease, the actual prevalence of disease at animal and holding level are uncertain. Previous attempts to estimate disease frequency at animal and flock-levels used data generated by surveillance before the eradication measures and the breeding for resistance programmes could have had any impact on the disease frequency. A prevalence estimate ranging from 0.33 to 2.06% using abattoir and fallen stock survey data for 2002 was obtained (Gubbins, 2008). A backcalculation approach was used to estimate the prevalence of sheep infected with classical scrapie, integrating data on reported clinical cases (1993–2007) and the results of fallen stock and abattoir surveys (2002–2007) (Gubbins and McIntyre, 2009). The authors reported an estimated prevalence of 0.6–0.7%, if infected animals could only be detected in the final quarter of the incubation period, and an overall decline between 2003 and 2007 of 40%, independent of the assumptions made about the diagnostic tests. In terms of holding prevalence, a minimum lower bound for the total number of holdings in GB infected with scrapie was estimated to be 642 (Del Rio Vilas et al., 2005) by applying capture–recapture methodology (CRC) to data from the abattoir and fallen stock surveys. Using data from the Compulsory Scrapie Flocks Scheme (CSFS) and the passive and active surveillance programmes between April 2005 and 2006, the number of scrapie-affected holdings in GB was estimated at around 350 (Del Rio Vilas and Böhning, 2008). The same authors applying one-list capture–recapture approaches to passive surveillance data for the years 2002, 2003 and 2004, estimate a flock prevalence of around 300 holdings per year (Del Rio Vilas and Böhning, 2008). No estimates of scrapie prevalence have been produced since 2007 and therefore there has been no attempt to assess the effectiveness of the breeding for resistance and the compulsory eradication measures in reducing the prevalence of scrapie in GB. The objective of this study is to estimate two measures of disease frequency using up to date surveillance data collected during and after the implementation of the different control measures established in 2004, and the breeding for resistance schemes that ran from 2001 until 2009. In particular, the two estimates produced were: (i) the animal prevalence (proportion of adult sheep infected with classical scrapie) in the national flock and (ii) the holding prevalence (percentage of holdings with at least an adult
243
animal infected with classical scrapie) in the national flock. 2. Methods 2.1. Model description: estimation of animal level prevalence by genotype group A back-calculation model was developed to integrate data from the abattoir survey and fallen stock surveys to determine the prevalence of infection in the national flock, taking into account the NSP genotype group of tested and positive sheep. The model is described in full as supplementary material (Appendix A). The model is similar in principle to a previous approach (Gubbins, 2008), except that (i) it estimates prevalence over several years data, (ii) it allows the rate of under-reporting of clinical suspects to vary over time, and (iii) it assumes that animals show overt clinical signs at clinical onset, whereas in (Gubbins, 2008) it is allowed for a large proportion of animals to die at clinical onset prior to overt clinical signs. 2.2. Estimation of the flock-level prevalence The flock-level prevalence of classical scrapie was estimated from the animal level prevalence, but adjusting for the clustering of infection within flocks (i.e. the withinflock prevalence distribution), and the effect of flock size on the likelihood of a flock being infected, p. The estimation of flock-level prevalence did not take genotype into account. The flock-level prevalence can be estimated from the estimator of animal-level prevalence, as the two are related, since the animal prevalence P(animal infected) is given by: P(animal inf ected) =
P(flock inf ected)
flocks
×P
within flock prevalence flock inf ected
The animal-level prevalence (P(animal infected)) was estimated from the 2012 data on scrapie positives, using the back-calculation approach (see supplementary material Appendix A). The within-flock prevalence distribution (P(within flock prevalence/flock infected)) was estimated using data from infected holdings culled and tested under the Compulsory Scrapie Flocks Scheme (CSFS) (Ortiz-Pelaez and Del Rio Vilas, 2009). Zero-truncated Poisson model and negative binomial models were fitted to the data on the number of positive sheep from the CSFS flocks with at least one positive sheep, in order to determine the per capita rate of scrapie infection. The per capita rate of infection from the CSFS data was multiplied by 2.2 to account for the ratio between detected positives and within-flock prevalence (Matthews et al., 2001). The term P(flock infected) was assumed be depend on holding size, as found in a previous study looking at the
244
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
effect of flock size on the likelihood of infection (Green et al., 2007) which showed an odds ratio of 2.69 per 1000 sheep. Therefore P(flock infected) was assumed to be of the form logit(˛ + ˇni ) with parameter ˇ derived from the odds ratio provided in Green et al. (2007), and ˛ determined so that the overall animal level prevalence was determined from the active surveillance data. The flock prevalence was then determined by finding the value of ˛ such that there was a match between the expected value of the animal prevalence and that estimated by the back-calculation model.
3. Input data In order to estimate the trend of scrapie infection prevalence, and to have sufficient data to provide estimates of the risk by genotype group, the back-calculation model was applied to annual data on the number of clinical suspects, fallen stock and abattoir survey data from GB between 2005 and 2012. Based on the level of resistance/susceptibility to classical scrapie, the 15 allelic variations at codons 136,154 and 171 of the ovine PrP gene present in GB breeds were grouped in five categories, from Type I to Type V.1 Observed age at onset of clinical cases between 1993 and 2012 stratified according to these five NSP types (I–V) was extracted from the Scrapie Notifications Database (SND), that contains data on all scrapie cases confirmed since scrapie became a notifiable disease in GB in 1993. A total of 4934 cases were included in the analysis between 1993 and 2011. Cases before 2003 (when atypical scrapie started to be discriminated) were considered all classical based on two criteria: the genotype of the animal susceptible to classical scrapie when available (as early as 1998) and/or the fact that the case was detected by passive surveillance, i.e., it was a clinical suspect. For a description of this database, see (Tongue et al., 2006). The number tested and positive animals in the abattoir and fallen stock surveys between 2005 and 2012 were extracted from the TSESS national database. This is the GB repository for scrapie active surveillance data including test results and epidemiologically associated data at animal level. The TSESS and SND databases are both maintained at the Animal Health and Veterinary Laboratories Agency (AHVLA). Number of positive and tested animals per year and surveillance route (abattoir survey and fallen stock) were extracted from the database and summarised in Tables D1 and D2 of Appendix D. The genotype distribution of the sheep population for each year between 2005 and 2012 was assumed to follow that of a sample of 8413 sheep tested for TSEs by the abattoir and the fallen stock surveys of January to December 2012 as the best possible
1 NSP genotype classification by degree of resistance/susceptibility to classical scrapie. In decreasing order of resistance: Type I (ARR/ARR) most resistant, Type II (ARR/AHQ, ARR/ARH, ARR/ARQ) resistant but need careful selection, Type III (AHQ/AHQ, AHQ/ARH, AHQ/ARQ, ARH/ARH, ARH/ARQ, ARQ/ARQ) little resistance, Type IV (ARR/VRQ) susceptible and Type V (AHQ/VRQ, ARH/VRQ, ARQ/VRQ, VRQ/VRQ) highly susceptible.
estimation available of the genotype distribution in the general population (Ortiz-Pelaez et al., 2014). 4. Model parameterisation 4.1. Incubation period distributions by NSP group The incubation period for scrapie was assumed to follow a lognormal distribution, in line with previous models of scrapie (e.g. Gubbins, 2008), and the lognormal distribution has also been found to be a good fit to incubation periods for other TSEs such as BSE (Wells et al., 2007). For NSP groups I–IV, the incubation period distribution was estimated separately for each by fitting a lognormal incubation period to the observed age at onset of clinical cases in the Scrapie Notifications Database (SND), accounting for the age distribution of the population (from Gubbins et al., 2003) (see Appendix B). For NSP group V, the ANOVA test showed significant differences between the log of the mean age of onset between some of the genotypes, with the mean of the logarithm of the age of onset of ARH/VRQ and VRQ/VRQ sheep being significantly different from that of AHQ/VRQ and ARQ/VRQ sheep. Therefore, the NSP group V incubation period was estimated separately for the following two groups: (i) ARH/VRQ (n = 110) and VRQ/VRQ (n = 414), and (ii) AHQ/VRQ (n = 3) and ARQ/VRQ (n = 1202). 4.2. Sensitivity of diagnostic test A key parameter of the model is the sensitivity of the rapid test to detect scrapie relative to the number of months before clinical onset. The rapid test used in GB is the Biorad TeSeE® test (sandwich ELISA using two monoclonal antibodies) (Bio-Rad Laboratories Ltd.), which although originally developed for cattle, has been found to have similar sensitivity to rapid tests approved for sheep (Bozetta et al., 2011). This parameter was estimated using data (AHVLA, unpublished) from the scrapie-infected flock kept by AHVLA, as described by Dexter et al. (2009) using the method developed for BSE (Arnold et al., 2007) (see Appendix C in supplementary material for full details). The key output of the statistical model by Arnold et al. (2007) was that the probability of detected PrPSc in an infected animal increases along with the incubation period, denoted (). The parameter () was given by the following logistic regression curve, commonly used to fit binary data: () =
exp(˛ + ˇ) 1 + exp(˛ + ˇ)
(1)
where ˛, ˇ were parameters estimated from the data. The statistical model of test sensitivity versus the proportion of the incubation period completed was fitted to the data from the AHVLA scrapie-infected flock for the VRQ/VRQ sheep, since this had the largest sample size (n = 269). It was assumed that sheep of other genotypes would have the same sensitivity versus the proportion of the incubation period completed. The specificity of the test was assumed to 100%, in line with findings from Philipp et al. (2005). The parameters of the logit model giving the sensitivity of the rapid test relative to the proportion of the incubation
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
245
period completed were ˛ = −11.96, ˇ = 17.2. This results in the estimated sensitivity (probability of detection) relative to the proportion of the incubation completed as shown in Fig. 1.
5. Model outputs The model provided estimates of the prevalence of infection and the number of infected animals in each NSP group. The model also estimated (i) the proportion of infected sheep entering the healthy slaughter stream, denoted K; (ii) the probability of a sheep being reported as a clinical suspect each year (see Appendix A). The relative risk of scrapie by NSP type was also estimated, for each of the types I–V (with no differentiation between the genotypes in group V, unlike for the incubation period estimation). Model fit for the numbers of animals positive in each surveillance stream was assessed by the chi square statistic, based on comparison of the case numbers in each stream each year between the modelled and observed counts, with 2008–2012 summed together to ensure sufficient counts
Fig. 1. The sensitivity of the rapid test for scrapie versus the proportion of the incubation period completed for VRQ/VRQ sheep, estimated from the Animal Health and Veterinary Laboratories Agency scrapie-infected flock (n = 269).
Fig. 2. Sum of the total scrapie cases in Great Britain for each year between 2005 and 2012 across all genotypes predicted by the model and that observed in the clinical suspect stream, the fallen stock survey and the abattoir survey. Sampling errors are included for the observed for the fallen stock and abattoir surveys, calculated from binomial confidence intervals.
246
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
Fig. 3. The estimated prevalence of sheep infected with scrapie in Great Britain between 2005 and 2012.
for the chi square to be valid. The number of infected sheep in each genotype group was estimated according to the method used in a previous study (Gubbins, 2008) (see Appendix A). 6. Results The total number of positive sheep in the abattoir and fallen stock surveys and those reported as clinical suspects between 2005 and 2012 in GB are shown in Fig. 2, along with the fit of the back-calculation model. The risks of classical scrapie infection in animals with genotypes of NSP Types I–IV, relative to Type V, were estimated to be: 0, 0.0008, 0.07, and 0.21, respectively. The model showed a good overall fit to the total cases each year (Fig. 2), with a chi square test showing no evidence of a lack of fit to the data (P = 0.17). Model estimates for the relative risk of infection and prevalence by NSP group are given in Table 1, which shows (as expected) an increasing risk of infection in successive NSP groups from Group I (0% prevalence) to Group V (0.64% prevalence). Assuming the total GB sheep population of 16.9 million (2011 Department for Environment, Food and Rural Affairs – Defra’s Sheep and Goat Inventory), and using Eq. (1) in Appendix A, the estimated number of sheep infected with classical scrapie is 5737 (0.034%) distributed as follows: Type I: 0; Type II: 42; Type III: 1425; Type IV: 952; Type V: 3318. Note that despite Type III having a much lower estimated prevalence, its greater contribution to the sheep national flock results in a similar number of animals infected than Type IV. According to the results of the back-calculation model, there was a significant exponential decline (Fig. 3), with a mean 31% reduction in prevalence year by year (95% CI: 0.27–0.37). This leads to a 90% drop in infection prevalence between 2005 and 2012. The model estimated a very low rate of reporting of clinical suspects, with estimates of 1.8% (95% CI: 1.6–2.1%) clinical cases reported for 2005 and 1.3% (95% CI: 1.1–1.6%) for 2006. A likelihood ratio test showed no significant difference between the reporting rates for each year from
Fig. 4. Comparison of the fit of the zero-truncated Poisson model for the observed number of scrapie positive sheep in flocks culled as part of the Compulsory Scrapie Flock Scheme in Great Britain.
2007 onwards (P = 0.12), so a common rate of reporting was fit for each year from 2007 to 2012. There was a large drop in the reporting rates for 2007–2012 compared with 2005 and 2006, with an estimate of the reporting rate 2007–2012 of 0.12% (95% CI: 0.07–0.21%), reflecting the decline from 412 reported clinical suspects in 2005 (179 of them confirmed), down to only a total of 39 clinical suspects and 7 cases confirmed in the 5 year period 2008–2012. Fitting a common rate of the parameter K (the proportion of infected sheep ending up in the healthy slaughter stream) throughout all years, resulted in an overestimate of the number of healthy slaughter positives in 2005. A likelihood ratio test showed a significant difference between the parameter K for 2005 and subsequent years (P < 0.01) and so a separate value of K was fitted for (i) 2005, K = 0.14 (95% CI: 0.07–0.26) and (ii) 2006–2012, K = 0.06 (95% CI: 0.04–0.06). A negative binomial distribution, formulated as a Poisson distribution allowing for over-dispersion and with a constant dispersion parameter, was found to fit the withinflock prevalence distribution in the CSFS data significantly better than a Poisson distribution (P < 0.001), as determined by a likelihood ratio test. The negative binomial distribution had parameters given by = 0.0034 (95% CI: −6.03 × 10−5 , 0.0068) (where the mean of the negative binomial=exp(ni ), where ni is the flock size) and ı = 1.69 (95% CI: 0.71–4.0), where ␦ determined the level of overdispersion. The fit of the model to the observed data is given in Fig. 4. The estimated holding level prevalence of classical scrapie for 2012 of 0.3% (95% CI: 0.05–0.61%). If the total number of sheep holdings in GB in 2012 (2012 Defra’s Sheep and Goat Inventory) was 72,218, then the number of expected infected sheep holdings with classical scrapie in 2012 is 215 (95% CI: 33–437). 7. Discussion The results of this study represent the first estimation of the burden of classical scrapie in the British sheep
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
247
Table 1 Summary of the estimates for the relative risk, the estimated prevalence and number infected by NSP genotype group from the back-calculation model of scrapie in GB, and the estimated proportion of each NSP genotype group in the GB sheep population. Genotype group
Relative risk of infection
Estimated proportion of population (%)
Estimated prevalence by NSP group (%)
Number infected by NSP group
I II III IV V
0 0.0008 0.07 0.21 1.0
28.8 43.8 20.7 3.6 3.1
0 0.0006 0.04 0.16 0.64
0 42 1425 952 3318
population after the implementation of the breeding for resistance as per the NSP schemes and the statutory control measures as per EU legislation. The former aimed to increase the frequency of the ARR allele and to reduce the frequency of highly susceptible alleles, especially VRQ, in high-genetic merit flocks (HGMF) with the view to disseminate these effects to the general population. Although not at the same scale as in the HGMF, there has been an absolute increase of 9% in the frequency of the ARR allele from 43.3% to 52.3% between 2002–2003 and 2012–2013 and 6.8% and 2.8% decreases in the frequencies of the ARQ and VRQ alleles, respectively (Ortiz-Pelaez et al., 2014). The latter decreased the infection pressure by removing infected animals and healthy animals with high-risk PrP genotypes and susceptible to acquire the disease in an infected environment. The findings of the present study are in agreement with those of Gubbins and McIntyre (2009), who estimated the scrapie infection prevalence in GB using a combination of data between 1993 and 2007, and reported a 40% decline between 2003 and 2007. The present study indicates that this decline continued from 2007, and the prevalence in 2012 has dropped to 10% of that in 2005. The present study estimate of 0.27% prevalence for 2005 in line with that of a previous study (Gubbins, 2008), which had an estimate of 0.98% with equivalent test sensitivity (100% in the last 25% of the incubation period) for 2002. There would have been a large reduction in infection prevalence between 2002 and 2005, as reflected by around 2.5 times as many reported cases in 2002 compared to 2005. The model is similar in principle to that of a previous back-calculation model to estimate scrapie prevalence from multiple data streams in GB (Gubbins, 2008). One important difference is the estimation of the differential slaughter risk (K). In the earlier study (Gubbins, 2008) this represents the combined increased risk of either being sent for slaughter prior to clinical signs or dying on farm prior to the onset of clinical signs. In the present study it was preferred to separate the risk of being sent for slaughter prior to clinical signs to that of dying of scrapie on farm prior to clinical onset. This may be the reason that the estimate of K in the present study (∼14% in 2005 and 6% thereafter) is much smaller than that found by Gubbins (2008) (55–92%, depending on assumptions). In the present study we chose to estimate the rate of under-reporting of scrapie suspects, rather than assuming a rate based on other studies, as in an earlier study (Gubbins, 2008). The rate for 2005 (1.8%) is much lower than the estimated 38% for 2002 by Gubbins (2008) and those reported from previous studies. Böhning and Del Rio Vilas (2008) quantified the completeness of identification of passive
surveillance as 45% and Sivam et al. (2006) estimated that more than 38% of the farmers that thought they had scrapie reported it to the veterinary authorities in 2002, based on responses to a postal survey to farmers. The main difference between the model in the previous study (Gubbins, 2008) and that of the present study, is that the former allows for sheep to die of scrapie prior to the onset of overt clinical signs. In order to match the observed number of reported cases and the data from the 2002 fallen stock and abattoir surveys, the model in the previous study (Gubbins, 2008) estimated that the majority (55–92%, depending on scenario) of scrapie infections surviving to disease onset are either sent for slaughter or die as fallen stock prior to the onset of clinical signs. In the model developed in the present study, animals reaching disease onset would be assumed to show clinical signs and would thus count as under-reported. A further potential difference in the rate of underreporting may be caused by differences in the specification of test sensitivity between the previous study (Gubbins, 2008) and the present one. In this regard, back-calculation models of scrapie are very sensitive to the estimate of the sensitivity of the rapid test relative to clinical onset. Previous studies have based their estimates on the age at detection from pathogenesis studies and have assumed ranges of detection at 3–6 months prior to clinical onset (Gubbins et al., 2003) or in the last 25 or 50% of the incubation period (Gubbins, 2008). The present study directly estimated the sensitivity of the test relative to disease onset, using maximum likelihood methods from experimental data from a timed cull. The results for the VRQ/VRQ genotype suggest that the test can detect animals with a greater than 50% probability in the last 25% of the incubation period, which is much more sensitive than the test for BSE (Arnold et al., 2007), and within the range of sensitivity relative to clinical onset considered in the previous study (Gubbins, 2008). However it remains uncertain whether the sensitivity of the test varies between genotypes. Limited availability of positive non-VRQ animals (64 ARQ/VRQ and 17 ARQ/ARQ) precluded the accurate estimation of the test sensitivity by genotype, and this fact could influence the estimates of animal level prevalence and risk by genotype group. However, recent findings from an experimental study with orally dosed sheep of several non-VRQ/VRQ genotypes has found only small differences in the timing of detection in CNS relative to the proportion of the incubation period completed for classical scrapie (L. Gonzalez and M. Jeffrey (AHVLA), unpublished observations) suggesting that the assumption of constant sensitivity between genotypes (versus the proportion
248
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
of the incubation period completed) may be appropriate. The large estimated decrease in the reporting rate between 2005 and 2007 onwards (from 1.8% to 0.12%) seems to be in excess of what would be expected due to the reduction of the scrapie prevalence alone. The initial compensation paid to owners of culled infected flocks in 2004 and 2005 exceeded in many cases the market value of the animals, providing an extra incentive to farmers to report clinical suspects, resulting in high levels of reporting (over 400 per year) (Böhning and Del Rio Vilas, 2008), in a very different approach to the one reported in the late nineties when farmers were more reluctant to report suspect cases because of the stigma associated with the disease (Hoinville et al., 1999). Furthermore, there could be a genotype effect on the rate of reporting. The reduction in the proportion of VRQ/VRQ sheep by half from 0.4% to 0.2% in 10 years (Ortiz-Pelaez, unpublished data) in the population as a result of the NSP will have resulted in an increase in the age at which sheep develop clinical scrapie, since the incubation period of VRQ/VRQ sheep is relatively short, and also potentially on the extent of the clinical signs; which may vary between genotypes, as has been found by previous studies (Healy et al., 2003; Konold et al., 2009). Recent changes in the genotype profile of the national flock may have influenced the degree of clinical signs shown by sheep. Reduction in the compensation could be one of the reasons in the sharp decline of reporting observed in 2006 and 2007 where 264 and 69 suspect cases were reported, respectively, and only 39 in total the 5-year period between 2008 and 2012. Defra and the Devolved Authorities (Scotland and Wales) published a revised version of the Scrapie Advisory Notes for Farmers in October 2009 (Anon, 2009). The Scrapie Advisory Notes are intended to assist sheep or goat keepers in fulfilling their legal obligations to immediately report to the veterinary authorities any animal that they suspect of being affected with scrapie. The notes are not designed to be a complete guide to a sheep or goat keeper’s obligations but a set of guidelines to encourage farmers to be vigilant and report suspect cases. Despite these efforts, scrapie has a low profile in the farming community and passive surveillance is no longer a source of new cases. With regards to the flock prevalence, the estimate calculated using an extension of the back-calculation model is considerably lower than previous estimates. However the different nature of the methods applied makes any comparison problematic. Capture–recapture methods rely on the comparison of lists of cases obtained from independent detection protocols (Brittain and Böhning, 2009), although the study of single case lists have also been applied, to estimate disease prevalence. For example, the first CRC model applied to three surveillance sources, one of which was SND (i.e. passive surveillance) and using 2002–2003 data reported an estimated number of infected holdings of 1653 (95% CI: 354–6438). The estimate reported in this study reflects a 90% reduction in the 10-year period, acknowledging the different approaches followed to produce such results, which is consistent with the large reduction in animal prevalence (Fig. 2). The estimate of 215 infected flocks is also broadly consistent with the estimation reported by
the same authors (Del Rio Vilas and Böhning, 2008) of 393 infected flocks (95% CI: 213–573) using a different data source: the CSFS 2005–2006, although the estimate from the present study would suggest only a 45% reduction from the 2005–2006 estimate of holding prevalence, compared to an estimated 90% reduction in prevalence at the animal level in that time period. Although small in number, the removal of heavily-infected flocks in the early years of the application of the compulsory control measures may have left a number of flocks with low prevalence of infection subdued by the increase in the overall genetic resistance at flock level. The holding prevalence estimate of 215 implies a relatively high average number of infected sheep per infected holding (5737 infected sheep in 215 holdings, indicates 26.7 infected sheep per holding). This is highly influenced by the degree of over-dispersion estimated from the CSFS data. This over-dispersion indicates a high degree of variability in the within-flock prevalence, and suggests the possibility of a few flocks having potentially large numbers of infected sheep. High prevalence flocks have been observed in GB, such as an infected large flock in which more than 200 cases were confirmed in 2011–2012 (Ortiz-Pelaez and Arnold, unpublished results). In one longitudinal study of 15 individual commercial infected sheep flocks, the flock-level prevalence estimates varied from 0 to 15.4% when culled animals were screened by immunohistochemistry for evidence of infection (Tongue et al., 2005). McIntyre et al. (2008) identified 415 cases of scrapie in 30 infected flocks, with numbers of cases per flock varying from 1 to 131, with seven flocks having only a single case of scrapie. The overall mean number of infected sheep per infected flock is further influenced by the greater likelihood that large flocks will be infected, with a significant influence of flock size on the probability of being infected (Green et al., 2007). If the implementation of the CSFS has influenced the within-flock prevalence distribution by removal of the high prevalence flocks, so that there is no longer such over-dispersion of the within-flock prevalence, this would have a significant effect on the estimate of holding-level prevalence. Repeating the estimation of holding prevalence but assuming a Poisson distribution in place of a negative binomial (and given the same mean within-flock prevalence of 2.2% apparent prevalence uprated by 2.2 for undetected infections) results in a holding prevalence of 404 flocks, almost double that assuming negative binomial within-flock prevalence. The estimate of flock prevalence also did not take into account the distribution of genotypes at the flock level i.e. there will be some flocks with a high proportion of scrapie resistant animals that are likely, if infected, to have a very low prevalence of within-flock infection, and other flocks with many susceptible animals with a high within flock prevalence. To incorporate the distribution of genotypes at the flock level, an analysis would need to account for the distribution of different flock genotype profiles in the national flock, for which there is limited data. Instead, the current approach relies on the estimate of within-flock prevalence from the cull of 121 CSFS flocks during the period 2005–2007, approximately half of all culled scrapieaffected flocks in GB.
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
The estimates of flock and animal prevalence confirm the decline of the burden of disease in GB and the combined effectiveness of the measures applied. It is difficult to apportion the decline in prevalence between each of the major control measures, the CSFS and the NSP. However, some insight may be gained from earlier predictions of the impact of breeding programmes on the animal prevalence in GB (Gubbins and Roden, 2006). Under the scenario where the proportion of sheep in NSP groups IV and V roughly halved (the scenario most similar to the observed decline in GB between 2002 and 2013, a drop from 12% to 7% (Ortiz-Pelaez et al., 2014; Appendix E)) Gubbins and Roden (2006) predicted an approximate halving of the prevalence of scrapie in GB. Assuming the modelling results from Gubbins and Roden (2006) are valid, this would suggest that the CSFS had a larger impact than the NSP on reducing sheep scrapie in GB, but that the NSP still had a significant effect. The estimates of animal and flock prevalence, while showing a large decline since 2005, also suggest that there are still a number of undetected flocks in GB with at least one infected animal. The delay in the detection due to the diminishing surveillance efforts, the relaxation of the control measures and the uncertainties on the current level of genetic resistance derived from the end of national schemes may all favour within-flock transmission leading to a higher prevalence in the remaining affected flocks than would have been expected at this stage of the epidemic (after nearly 10 years of enforcing the control measures and 8 years of breeding for resistance) given the within-flock prevalence distribution from the CSFS flocks (Fig. 4). This was showed in a large outbreak detected incidentally in a large flock where a total of 122 cases were confirmed in 2011, making it one of the largest outbreaks of classical scrapie in a single holding ever detected in GB (Ortiz-Pelaez et al., 2012). Further studies on the ability of the current active surveillance streams to detect scrapie infected animals/flocks at the reported prevalence levels and on the cost associated to them would be advisable.
8. Conclusions The decline in detection of scrapie cases continued from 2007, and this present study shows the prevalence in 2012 has dropped to 10% of that in 2005 indicating the effectiveness of the control measures. It also shows a bias in the destination of infected animals, with the majority of infected animals being detected in the fallen stock stream, and an extremely low proportion of animals detected as clinical suspects; this is very important in terms of the design of surveillance schemes for classical scrapie.
Acknowledgements This study was funded by the Department for Environment, Food and Rural Affairs (Defra) under the project TS5980. The helpful comments of two anonymous referees is also gratefully acknowledged.
249
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.prevetmed.2014.07.015. References Anon, 2009. Scrapie—advisory notes for farmers, http://archive.defra. gov.uk/foodfarm/farmanimal/diseases/atoz/bse/documents/scrapeng.pdf (accessed 21.03.14). Arnold, M.E., Ryan, J.B.M., Konold, T., Simmons, M.M., Spencer, Y.I., Wear, A., Chaplin, M., Stack, M., Czub, S., Mueller, R., Webb, P.R., Davis, A., Spiropoulos, J., Holdaway, J., Hawkins, S.A.C., Austin, A.R., Well s, G.A.H., 2007. Estimating the temporal relationship between PrPSc detection and incubation period in experimental bovine spongiform encephalopathy of cattle. J. Gen. Virol. 88, 3198–3208. Böhning, D., Del Rio Vilas, V.J., 2008. Estimating the hidden number of scrapie affected holdings in Great Britain using a simple, truncated count model Allowing for Heterogeneity. J. Agric. Biol. Environ. Stat. 13, 1–22. Bozetta, E., Nappi, R., Crudeli, S., Meloni, D., Varello, K., Loprevite, D., Melis, P.G., Mazza, M., Colussi, S., Ingravalle, F., Ru, G., Nonno, R., Ligios, C., 2011. Comparative performance of three TSE rapid tests for surveillance in healthy sheep affected by scrapie. J. Virol. Methods 173, 161–168. Brittain, S., Böhning, D., 2009. Estimators in capture recapture with two sources. Adv. Stat. Anal. 93, 23–47. Del Rio Vilas, V.J., Sayers, R., Sivam, K., Pfeiffer, D., Guitian, J., Wilesmith, J.W., 2005. A case study of capture–recapture methodology using scrapie surveillance data in Great Britain. Prev. Vet. Med. 67, 303–317. Del Rio Vilas, V.J., Böhning, D., 2008. Application of one-list capture–recapture models to scrapie surveillance data in Great Britain. Prev. Vet. Med. 85, 253–266. Dexter, G., Tongue, S.C., Heasman, L., Bellworthy, S.J., Davis, A., Moore, S.J., Simmons, M.M., Sayers, A.R., Simmons, H.A., Matthews, D., 2009. The evaluation of exposure risks for natural transmission of scrapie within an infected flock. BMC Vet. Res. 5, 38. Green, D.M., Del Rio Vilas, V.J., Birch, C.P.D., Johnson, J., Kiss, I.Z., McCarthy, N.D., Kao, R.R., 2007. Demographic risk factors for classical and atypical scrapie in Great Britain. J. Gen. Virol. 207, 3486–3492. Gubbins, S., Simmons, M.M., Sivam, K., Webb, C.R., Hoinville, L.J., 2003. Prevalence of scrapie infection in Great Britain: interpreting the results of the 1997–1998 abattoir survey. Proc. R. Soc. B. 270, 1919–1924. Gubbins, S., Roden, J.A., 2006. Breeding programmes for TSE resistance in British sheep II. Assessing the impact on the prevalence and incidence of scrapie. Prev. Vet. Med. 73, 17–31. Gubbins, S., 2008. Prevalence of sheep infected with classical scrapie in Great Britain: integrating multiple sources of surveillance data for 2002. J. R. Soc. Interface 5, 1343–1351. Gubbins, S., McIntyre, M., 2009. Prevalence of sheep infected with classical scrapie in Great Britain, 1993–2007. Epidemiol. Inf. 137, 787–791. Healy, A.M., Weavers, E., McElroy, M., Gomez-Parada, M., Collins, J.D., O’Doherty, E., Sweeney, T., Doherty, M.L., 2003. The clinical neurology of scrapie in Irish sheep. J. Vet. Intern. Med. 17, 908–916. Hoinville, L., McLean, A.R., Hoek, A., Gravenor, M.B., Wilesmith, J., 1999. Scrapie occurrence in Great Britain. Vet. Rec. 145, 405–406. Konold, T., Bone, G., Ortiz-Pelaez, A., Tortosa, R., Clifford, D., Dexter, G., Simmons, M.M., Spiropoulos, J., Berthelin-Baker, C.F., 2009. Associations of clinical signs and prion protein genotypes in British sheep with scrapie. Dtsch. Tierärztl. Wochenschr. 116, 380–388. Matthews, L., Coen, P.G., Foster, J.D., Hunter, N., Woolhouse, M.E.J., 2001. Population dynamics of a scrapie outbreak. Arch. Virol. 146, 1173–1186. McIntyre, K.M., Gubbins, S., Goldmann, W., Hunter, N., Baylis, M., 2008. Epidemiological characteristics of classical scrapie outbreaks in 30 sheep flocks in the United Kingdom. PLoS ONE 3, e3994. Ortiz-Pelaez, Á., Del Rio Vilas, V., 2009. Within-holding prevalence of sheep classical scrapie in Great Britain. BMC Vet. Res. 5, 1. Ortiz-Pelaez, A., Arnold, M.E., Warner, R., 2012. Sheep and goats scrapie surveillance 2011. Joint Descriptive Report for Great Britain, 2012, http://www.defra.gov.uk/ahvla-en/files/pub-tse-stats-scrapie.pdf
250
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
Ortiz-Pelaez, A., Thompson, C.E., Dawson, M., 2014. The impact of the National Scrapie Plan on the PRNP genotype distribution of the British national flock, 2002–2012. Vet. Rec., http://dx.doi.org/ 10.1136/vr.102255. Philipp, W.J., Kollmorgen, N., van Iwaarden, P., Charout-Got, J., Contreras Lopez, M., Goll, M., LeGuern, L., Schimmel, H., Vodrazka, P., 2005. The evaluation of rapid post mortem tests for the diagnosis of TSE in sheep. In: Report by the Institute for Reference Materials and Methods 2005. European Commission, See http://irmm.jrc.ec.europa.eu/ activities/TSE testing/Documents/040616 scrapie report global.pdf (accessed 21.03.14). Sivam, S.K., Baylis, M., Gravenor, M.B., Gubbins, S., 2006. Descriptive analysis of the results of an anonymous postal survey of the occurrence of
scrapie in Great Britain in 2002. Vet. Rec. 158, 501–506. Tongue, S.C., Webb, P., Simmons, M.M., Gubbins, S., 2005. Prevalence of scrapie infection in cull animals from 14 scrapie-affected flocks in Great Britain. Vet. Rec. 157, 480–482. Tongue, S.C., Pfeiffer, D.U., Warner, R., Elliott, H., Del Rio Vilas, V., 2006. Estimation of the relative risk of developing clinical scrapie: the role of prion protein (PrP) genotype and selection bias. Vet. Rec. 158, 43–50. Wells, G.A.H., Konold, T., Arnold, M.E., Austin, A.R., Hawkins, S.A.C., Stack, M., Simmons, M.M., Lee, Y.H., Gavier-Widén, D., Dawson, M., Wilesmith, J.W., 2007. Bovine spongiform encephalopathy: the effect of oral exposure dose on attack rate and incubation period in cattle. J. Gen. Virol. 88, 1363–1373.
Contents lists available at ScienceDirect
Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed
The evolution of the prevalence of classical scrapie in sheep in Great Britain using surveillance data between 2005 and 2012 Mark Arnold a,∗ , Angel Ortiz-Pelaez b a
Animal Health and Veterinary Laboratories Agency (AHVLA), The Elms, College Road, Sutton Bonington, Loughborough, LE12 5RB, UK Epidemiology, Surveillance and Risk Group, Animal Health and Veterinary Laboratories Agency, New Haw, Addlestone, Surrey, KT15 3NB, UK b
a r t i c l e
i n f o
Article history: Received 7 October 2013 Received in revised form 11 July 2014 Accepted 31 July 2014 Keywords: Scrapie Back-calculation methods Prevalence estimation Maximum likelihood methods
a b s t r a c t After the decline of the Bovine Spongiform Encephalopathy (BSE) epidemic in Great Britain (GB), scrapie remains the most prevalent animal Transmissible Spongiform Encephalopathy (TSE) present in GB. A number of control measures have been implemented for classical scrapie, and since 2005 there has been a large reduction in the number of observed cases. The objective of this study is to estimate two measures of disease frequency using up to date surveillance data collected during and after the implementation of different control measures established since 2004, and breeding for resistance schemes that ran from 2001 until 2009. This would enable an assessment of the effectiveness of both the breeding for resistance programme and the compulsory eradication measures in reducing the prevalence of scrapie in GB. Evaluation of the sensitivity of the rapid post-mortem test for scrapie indicated that it detected scrapie in the last 25% of the incubation period. A back-calculation model was developed to estimate the prevalence of infection at animal and flock-level. The results of the model indicated a mean drop of infection prevalence of 31% each year, leading to a 90% drop in infection prevalence between 2005, with an estimate of 5737 infected sheep in GB in 2012. The risks of classical scrapie infection in animals with genotypes of National Scrapie Plan Types I–IV (all other genotypes), relative to Type V (all genotypes containing V136 R154 Q171 and not A136 R154 R171 ), were estimated to be: 0, 0.0008, 0.07, and 0.21 respectively. The model estimated a very low rate of reporting of clinical suspects and a large decline from 2007 of the probability of a sheep being reported as a clinical suspect. The model also estimated that the expected number of sheep holdings with classical scrapie in 2012 was 215 (95% confidence interval: 33–437), out of a total of approximately 72,000 sheep holdings in GB. Model estimates indicate that the prevalence in 2012 has dropped to 10% of that in 2005, showing the effectiveness of the control measures. It also shows a bias in the destination of infected animals, with the majority of infected animals being detected in the fallen stock surveillance stream, and an extremely low proportion of animals detected as clinical suspects; this is very important in terms of the design of surveillance schemes for classical scrapie. Crown Copyright © 2014 Published by Elsevier B.V. All rights reserved.
1. Background ∗ Corresponding author. Tel.: +44 01509 678340; fax: +44 01509 674805. E-mail address: [email protected] (M. Arnold).
After the decline of the Bovine Spongiform Encephalopathy (BSE) epidemic in Great Britain (GB), scrapie remains the most prevalent animal Transmissible Spongiform
http://dx.doi.org/10.1016/j.prevetmed.2014.07.015 0167-5877/Crown Copyright © 2014 Published by Elsevier B.V. All rights reserved.
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
Encephalopathy (TSE) present in the country. The number of clinical scrapie suspects notified by farmers has drastically declined in the last 6 years. Since the disease became a notifiable disease in 1993, 2010 and 2012 have been the only 2 years where no clinical cases of scrapie were confirmed in GB and only three clinical index cases of classical scrapie were confirmed in GB in 2009, and the same number in 2011. The number of clinical cases and the apparent prevalence in the different surveillance streams have shown a continuous decline of classical scrapie over the last 6 years, although no statistically significant differences have been detected year by year (Ortiz-Pelaez et al., 2012). Although surveillance data provide a trend on the evolution of the incidence of the disease, the actual prevalence of disease at animal and holding level are uncertain. Previous attempts to estimate disease frequency at animal and flock-levels used data generated by surveillance before the eradication measures and the breeding for resistance programmes could have had any impact on the disease frequency. A prevalence estimate ranging from 0.33 to 2.06% using abattoir and fallen stock survey data for 2002 was obtained (Gubbins, 2008). A backcalculation approach was used to estimate the prevalence of sheep infected with classical scrapie, integrating data on reported clinical cases (1993–2007) and the results of fallen stock and abattoir surveys (2002–2007) (Gubbins and McIntyre, 2009). The authors reported an estimated prevalence of 0.6–0.7%, if infected animals could only be detected in the final quarter of the incubation period, and an overall decline between 2003 and 2007 of 40%, independent of the assumptions made about the diagnostic tests. In terms of holding prevalence, a minimum lower bound for the total number of holdings in GB infected with scrapie was estimated to be 642 (Del Rio Vilas et al., 2005) by applying capture–recapture methodology (CRC) to data from the abattoir and fallen stock surveys. Using data from the Compulsory Scrapie Flocks Scheme (CSFS) and the passive and active surveillance programmes between April 2005 and 2006, the number of scrapie-affected holdings in GB was estimated at around 350 (Del Rio Vilas and Böhning, 2008). The same authors applying one-list capture–recapture approaches to passive surveillance data for the years 2002, 2003 and 2004, estimate a flock prevalence of around 300 holdings per year (Del Rio Vilas and Böhning, 2008). No estimates of scrapie prevalence have been produced since 2007 and therefore there has been no attempt to assess the effectiveness of the breeding for resistance and the compulsory eradication measures in reducing the prevalence of scrapie in GB. The objective of this study is to estimate two measures of disease frequency using up to date surveillance data collected during and after the implementation of the different control measures established in 2004, and the breeding for resistance schemes that ran from 2001 until 2009. In particular, the two estimates produced were: (i) the animal prevalence (proportion of adult sheep infected with classical scrapie) in the national flock and (ii) the holding prevalence (percentage of holdings with at least an adult
243
animal infected with classical scrapie) in the national flock. 2. Methods 2.1. Model description: estimation of animal level prevalence by genotype group A back-calculation model was developed to integrate data from the abattoir survey and fallen stock surveys to determine the prevalence of infection in the national flock, taking into account the NSP genotype group of tested and positive sheep. The model is described in full as supplementary material (Appendix A). The model is similar in principle to a previous approach (Gubbins, 2008), except that (i) it estimates prevalence over several years data, (ii) it allows the rate of under-reporting of clinical suspects to vary over time, and (iii) it assumes that animals show overt clinical signs at clinical onset, whereas in (Gubbins, 2008) it is allowed for a large proportion of animals to die at clinical onset prior to overt clinical signs. 2.2. Estimation of the flock-level prevalence The flock-level prevalence of classical scrapie was estimated from the animal level prevalence, but adjusting for the clustering of infection within flocks (i.e. the withinflock prevalence distribution), and the effect of flock size on the likelihood of a flock being infected, p. The estimation of flock-level prevalence did not take genotype into account. The flock-level prevalence can be estimated from the estimator of animal-level prevalence, as the two are related, since the animal prevalence P(animal infected) is given by: P(animal inf ected) =
P(flock inf ected)
flocks
×P
within flock prevalence flock inf ected
The animal-level prevalence (P(animal infected)) was estimated from the 2012 data on scrapie positives, using the back-calculation approach (see supplementary material Appendix A). The within-flock prevalence distribution (P(within flock prevalence/flock infected)) was estimated using data from infected holdings culled and tested under the Compulsory Scrapie Flocks Scheme (CSFS) (Ortiz-Pelaez and Del Rio Vilas, 2009). Zero-truncated Poisson model and negative binomial models were fitted to the data on the number of positive sheep from the CSFS flocks with at least one positive sheep, in order to determine the per capita rate of scrapie infection. The per capita rate of infection from the CSFS data was multiplied by 2.2 to account for the ratio between detected positives and within-flock prevalence (Matthews et al., 2001). The term P(flock infected) was assumed be depend on holding size, as found in a previous study looking at the
244
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
effect of flock size on the likelihood of infection (Green et al., 2007) which showed an odds ratio of 2.69 per 1000 sheep. Therefore P(flock infected) was assumed to be of the form logit(˛ + ˇni ) with parameter ˇ derived from the odds ratio provided in Green et al. (2007), and ˛ determined so that the overall animal level prevalence was determined from the active surveillance data. The flock prevalence was then determined by finding the value of ˛ such that there was a match between the expected value of the animal prevalence and that estimated by the back-calculation model.
3. Input data In order to estimate the trend of scrapie infection prevalence, and to have sufficient data to provide estimates of the risk by genotype group, the back-calculation model was applied to annual data on the number of clinical suspects, fallen stock and abattoir survey data from GB between 2005 and 2012. Based on the level of resistance/susceptibility to classical scrapie, the 15 allelic variations at codons 136,154 and 171 of the ovine PrP gene present in GB breeds were grouped in five categories, from Type I to Type V.1 Observed age at onset of clinical cases between 1993 and 2012 stratified according to these five NSP types (I–V) was extracted from the Scrapie Notifications Database (SND), that contains data on all scrapie cases confirmed since scrapie became a notifiable disease in GB in 1993. A total of 4934 cases were included in the analysis between 1993 and 2011. Cases before 2003 (when atypical scrapie started to be discriminated) were considered all classical based on two criteria: the genotype of the animal susceptible to classical scrapie when available (as early as 1998) and/or the fact that the case was detected by passive surveillance, i.e., it was a clinical suspect. For a description of this database, see (Tongue et al., 2006). The number tested and positive animals in the abattoir and fallen stock surveys between 2005 and 2012 were extracted from the TSESS national database. This is the GB repository for scrapie active surveillance data including test results and epidemiologically associated data at animal level. The TSESS and SND databases are both maintained at the Animal Health and Veterinary Laboratories Agency (AHVLA). Number of positive and tested animals per year and surveillance route (abattoir survey and fallen stock) were extracted from the database and summarised in Tables D1 and D2 of Appendix D. The genotype distribution of the sheep population for each year between 2005 and 2012 was assumed to follow that of a sample of 8413 sheep tested for TSEs by the abattoir and the fallen stock surveys of January to December 2012 as the best possible
1 NSP genotype classification by degree of resistance/susceptibility to classical scrapie. In decreasing order of resistance: Type I (ARR/ARR) most resistant, Type II (ARR/AHQ, ARR/ARH, ARR/ARQ) resistant but need careful selection, Type III (AHQ/AHQ, AHQ/ARH, AHQ/ARQ, ARH/ARH, ARH/ARQ, ARQ/ARQ) little resistance, Type IV (ARR/VRQ) susceptible and Type V (AHQ/VRQ, ARH/VRQ, ARQ/VRQ, VRQ/VRQ) highly susceptible.
estimation available of the genotype distribution in the general population (Ortiz-Pelaez et al., 2014). 4. Model parameterisation 4.1. Incubation period distributions by NSP group The incubation period for scrapie was assumed to follow a lognormal distribution, in line with previous models of scrapie (e.g. Gubbins, 2008), and the lognormal distribution has also been found to be a good fit to incubation periods for other TSEs such as BSE (Wells et al., 2007). For NSP groups I–IV, the incubation period distribution was estimated separately for each by fitting a lognormal incubation period to the observed age at onset of clinical cases in the Scrapie Notifications Database (SND), accounting for the age distribution of the population (from Gubbins et al., 2003) (see Appendix B). For NSP group V, the ANOVA test showed significant differences between the log of the mean age of onset between some of the genotypes, with the mean of the logarithm of the age of onset of ARH/VRQ and VRQ/VRQ sheep being significantly different from that of AHQ/VRQ and ARQ/VRQ sheep. Therefore, the NSP group V incubation period was estimated separately for the following two groups: (i) ARH/VRQ (n = 110) and VRQ/VRQ (n = 414), and (ii) AHQ/VRQ (n = 3) and ARQ/VRQ (n = 1202). 4.2. Sensitivity of diagnostic test A key parameter of the model is the sensitivity of the rapid test to detect scrapie relative to the number of months before clinical onset. The rapid test used in GB is the Biorad TeSeE® test (sandwich ELISA using two monoclonal antibodies) (Bio-Rad Laboratories Ltd.), which although originally developed for cattle, has been found to have similar sensitivity to rapid tests approved for sheep (Bozetta et al., 2011). This parameter was estimated using data (AHVLA, unpublished) from the scrapie-infected flock kept by AHVLA, as described by Dexter et al. (2009) using the method developed for BSE (Arnold et al., 2007) (see Appendix C in supplementary material for full details). The key output of the statistical model by Arnold et al. (2007) was that the probability of detected PrPSc in an infected animal increases along with the incubation period, denoted (). The parameter () was given by the following logistic regression curve, commonly used to fit binary data: () =
exp(˛ + ˇ) 1 + exp(˛ + ˇ)
(1)
where ˛, ˇ were parameters estimated from the data. The statistical model of test sensitivity versus the proportion of the incubation period completed was fitted to the data from the AHVLA scrapie-infected flock for the VRQ/VRQ sheep, since this had the largest sample size (n = 269). It was assumed that sheep of other genotypes would have the same sensitivity versus the proportion of the incubation period completed. The specificity of the test was assumed to 100%, in line with findings from Philipp et al. (2005). The parameters of the logit model giving the sensitivity of the rapid test relative to the proportion of the incubation
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
245
period completed were ˛ = −11.96, ˇ = 17.2. This results in the estimated sensitivity (probability of detection) relative to the proportion of the incubation completed as shown in Fig. 1.
5. Model outputs The model provided estimates of the prevalence of infection and the number of infected animals in each NSP group. The model also estimated (i) the proportion of infected sheep entering the healthy slaughter stream, denoted K; (ii) the probability of a sheep being reported as a clinical suspect each year (see Appendix A). The relative risk of scrapie by NSP type was also estimated, for each of the types I–V (with no differentiation between the genotypes in group V, unlike for the incubation period estimation). Model fit for the numbers of animals positive in each surveillance stream was assessed by the chi square statistic, based on comparison of the case numbers in each stream each year between the modelled and observed counts, with 2008–2012 summed together to ensure sufficient counts
Fig. 1. The sensitivity of the rapid test for scrapie versus the proportion of the incubation period completed for VRQ/VRQ sheep, estimated from the Animal Health and Veterinary Laboratories Agency scrapie-infected flock (n = 269).
Fig. 2. Sum of the total scrapie cases in Great Britain for each year between 2005 and 2012 across all genotypes predicted by the model and that observed in the clinical suspect stream, the fallen stock survey and the abattoir survey. Sampling errors are included for the observed for the fallen stock and abattoir surveys, calculated from binomial confidence intervals.
246
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
Fig. 3. The estimated prevalence of sheep infected with scrapie in Great Britain between 2005 and 2012.
for the chi square to be valid. The number of infected sheep in each genotype group was estimated according to the method used in a previous study (Gubbins, 2008) (see Appendix A). 6. Results The total number of positive sheep in the abattoir and fallen stock surveys and those reported as clinical suspects between 2005 and 2012 in GB are shown in Fig. 2, along with the fit of the back-calculation model. The risks of classical scrapie infection in animals with genotypes of NSP Types I–IV, relative to Type V, were estimated to be: 0, 0.0008, 0.07, and 0.21, respectively. The model showed a good overall fit to the total cases each year (Fig. 2), with a chi square test showing no evidence of a lack of fit to the data (P = 0.17). Model estimates for the relative risk of infection and prevalence by NSP group are given in Table 1, which shows (as expected) an increasing risk of infection in successive NSP groups from Group I (0% prevalence) to Group V (0.64% prevalence). Assuming the total GB sheep population of 16.9 million (2011 Department for Environment, Food and Rural Affairs – Defra’s Sheep and Goat Inventory), and using Eq. (1) in Appendix A, the estimated number of sheep infected with classical scrapie is 5737 (0.034%) distributed as follows: Type I: 0; Type II: 42; Type III: 1425; Type IV: 952; Type V: 3318. Note that despite Type III having a much lower estimated prevalence, its greater contribution to the sheep national flock results in a similar number of animals infected than Type IV. According to the results of the back-calculation model, there was a significant exponential decline (Fig. 3), with a mean 31% reduction in prevalence year by year (95% CI: 0.27–0.37). This leads to a 90% drop in infection prevalence between 2005 and 2012. The model estimated a very low rate of reporting of clinical suspects, with estimates of 1.8% (95% CI: 1.6–2.1%) clinical cases reported for 2005 and 1.3% (95% CI: 1.1–1.6%) for 2006. A likelihood ratio test showed no significant difference between the reporting rates for each year from
Fig. 4. Comparison of the fit of the zero-truncated Poisson model for the observed number of scrapie positive sheep in flocks culled as part of the Compulsory Scrapie Flock Scheme in Great Britain.
2007 onwards (P = 0.12), so a common rate of reporting was fit for each year from 2007 to 2012. There was a large drop in the reporting rates for 2007–2012 compared with 2005 and 2006, with an estimate of the reporting rate 2007–2012 of 0.12% (95% CI: 0.07–0.21%), reflecting the decline from 412 reported clinical suspects in 2005 (179 of them confirmed), down to only a total of 39 clinical suspects and 7 cases confirmed in the 5 year period 2008–2012. Fitting a common rate of the parameter K (the proportion of infected sheep ending up in the healthy slaughter stream) throughout all years, resulted in an overestimate of the number of healthy slaughter positives in 2005. A likelihood ratio test showed a significant difference between the parameter K for 2005 and subsequent years (P < 0.01) and so a separate value of K was fitted for (i) 2005, K = 0.14 (95% CI: 0.07–0.26) and (ii) 2006–2012, K = 0.06 (95% CI: 0.04–0.06). A negative binomial distribution, formulated as a Poisson distribution allowing for over-dispersion and with a constant dispersion parameter, was found to fit the withinflock prevalence distribution in the CSFS data significantly better than a Poisson distribution (P < 0.001), as determined by a likelihood ratio test. The negative binomial distribution had parameters given by = 0.0034 (95% CI: −6.03 × 10−5 , 0.0068) (where the mean of the negative binomial=exp(ni ), where ni is the flock size) and ı = 1.69 (95% CI: 0.71–4.0), where ␦ determined the level of overdispersion. The fit of the model to the observed data is given in Fig. 4. The estimated holding level prevalence of classical scrapie for 2012 of 0.3% (95% CI: 0.05–0.61%). If the total number of sheep holdings in GB in 2012 (2012 Defra’s Sheep and Goat Inventory) was 72,218, then the number of expected infected sheep holdings with classical scrapie in 2012 is 215 (95% CI: 33–437). 7. Discussion The results of this study represent the first estimation of the burden of classical scrapie in the British sheep
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
247
Table 1 Summary of the estimates for the relative risk, the estimated prevalence and number infected by NSP genotype group from the back-calculation model of scrapie in GB, and the estimated proportion of each NSP genotype group in the GB sheep population. Genotype group
Relative risk of infection
Estimated proportion of population (%)
Estimated prevalence by NSP group (%)
Number infected by NSP group
I II III IV V
0 0.0008 0.07 0.21 1.0
28.8 43.8 20.7 3.6 3.1
0 0.0006 0.04 0.16 0.64
0 42 1425 952 3318
population after the implementation of the breeding for resistance as per the NSP schemes and the statutory control measures as per EU legislation. The former aimed to increase the frequency of the ARR allele and to reduce the frequency of highly susceptible alleles, especially VRQ, in high-genetic merit flocks (HGMF) with the view to disseminate these effects to the general population. Although not at the same scale as in the HGMF, there has been an absolute increase of 9% in the frequency of the ARR allele from 43.3% to 52.3% between 2002–2003 and 2012–2013 and 6.8% and 2.8% decreases in the frequencies of the ARQ and VRQ alleles, respectively (Ortiz-Pelaez et al., 2014). The latter decreased the infection pressure by removing infected animals and healthy animals with high-risk PrP genotypes and susceptible to acquire the disease in an infected environment. The findings of the present study are in agreement with those of Gubbins and McIntyre (2009), who estimated the scrapie infection prevalence in GB using a combination of data between 1993 and 2007, and reported a 40% decline between 2003 and 2007. The present study indicates that this decline continued from 2007, and the prevalence in 2012 has dropped to 10% of that in 2005. The present study estimate of 0.27% prevalence for 2005 in line with that of a previous study (Gubbins, 2008), which had an estimate of 0.98% with equivalent test sensitivity (100% in the last 25% of the incubation period) for 2002. There would have been a large reduction in infection prevalence between 2002 and 2005, as reflected by around 2.5 times as many reported cases in 2002 compared to 2005. The model is similar in principle to that of a previous back-calculation model to estimate scrapie prevalence from multiple data streams in GB (Gubbins, 2008). One important difference is the estimation of the differential slaughter risk (K). In the earlier study (Gubbins, 2008) this represents the combined increased risk of either being sent for slaughter prior to clinical signs or dying on farm prior to the onset of clinical signs. In the present study it was preferred to separate the risk of being sent for slaughter prior to clinical signs to that of dying of scrapie on farm prior to clinical onset. This may be the reason that the estimate of K in the present study (∼14% in 2005 and 6% thereafter) is much smaller than that found by Gubbins (2008) (55–92%, depending on assumptions). In the present study we chose to estimate the rate of under-reporting of scrapie suspects, rather than assuming a rate based on other studies, as in an earlier study (Gubbins, 2008). The rate for 2005 (1.8%) is much lower than the estimated 38% for 2002 by Gubbins (2008) and those reported from previous studies. Böhning and Del Rio Vilas (2008) quantified the completeness of identification of passive
surveillance as 45% and Sivam et al. (2006) estimated that more than 38% of the farmers that thought they had scrapie reported it to the veterinary authorities in 2002, based on responses to a postal survey to farmers. The main difference between the model in the previous study (Gubbins, 2008) and that of the present study, is that the former allows for sheep to die of scrapie prior to the onset of overt clinical signs. In order to match the observed number of reported cases and the data from the 2002 fallen stock and abattoir surveys, the model in the previous study (Gubbins, 2008) estimated that the majority (55–92%, depending on scenario) of scrapie infections surviving to disease onset are either sent for slaughter or die as fallen stock prior to the onset of clinical signs. In the model developed in the present study, animals reaching disease onset would be assumed to show clinical signs and would thus count as under-reported. A further potential difference in the rate of underreporting may be caused by differences in the specification of test sensitivity between the previous study (Gubbins, 2008) and the present one. In this regard, back-calculation models of scrapie are very sensitive to the estimate of the sensitivity of the rapid test relative to clinical onset. Previous studies have based their estimates on the age at detection from pathogenesis studies and have assumed ranges of detection at 3–6 months prior to clinical onset (Gubbins et al., 2003) or in the last 25 or 50% of the incubation period (Gubbins, 2008). The present study directly estimated the sensitivity of the test relative to disease onset, using maximum likelihood methods from experimental data from a timed cull. The results for the VRQ/VRQ genotype suggest that the test can detect animals with a greater than 50% probability in the last 25% of the incubation period, which is much more sensitive than the test for BSE (Arnold et al., 2007), and within the range of sensitivity relative to clinical onset considered in the previous study (Gubbins, 2008). However it remains uncertain whether the sensitivity of the test varies between genotypes. Limited availability of positive non-VRQ animals (64 ARQ/VRQ and 17 ARQ/ARQ) precluded the accurate estimation of the test sensitivity by genotype, and this fact could influence the estimates of animal level prevalence and risk by genotype group. However, recent findings from an experimental study with orally dosed sheep of several non-VRQ/VRQ genotypes has found only small differences in the timing of detection in CNS relative to the proportion of the incubation period completed for classical scrapie (L. Gonzalez and M. Jeffrey (AHVLA), unpublished observations) suggesting that the assumption of constant sensitivity between genotypes (versus the proportion
248
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
of the incubation period completed) may be appropriate. The large estimated decrease in the reporting rate between 2005 and 2007 onwards (from 1.8% to 0.12%) seems to be in excess of what would be expected due to the reduction of the scrapie prevalence alone. The initial compensation paid to owners of culled infected flocks in 2004 and 2005 exceeded in many cases the market value of the animals, providing an extra incentive to farmers to report clinical suspects, resulting in high levels of reporting (over 400 per year) (Böhning and Del Rio Vilas, 2008), in a very different approach to the one reported in the late nineties when farmers were more reluctant to report suspect cases because of the stigma associated with the disease (Hoinville et al., 1999). Furthermore, there could be a genotype effect on the rate of reporting. The reduction in the proportion of VRQ/VRQ sheep by half from 0.4% to 0.2% in 10 years (Ortiz-Pelaez, unpublished data) in the population as a result of the NSP will have resulted in an increase in the age at which sheep develop clinical scrapie, since the incubation period of VRQ/VRQ sheep is relatively short, and also potentially on the extent of the clinical signs; which may vary between genotypes, as has been found by previous studies (Healy et al., 2003; Konold et al., 2009). Recent changes in the genotype profile of the national flock may have influenced the degree of clinical signs shown by sheep. Reduction in the compensation could be one of the reasons in the sharp decline of reporting observed in 2006 and 2007 where 264 and 69 suspect cases were reported, respectively, and only 39 in total the 5-year period between 2008 and 2012. Defra and the Devolved Authorities (Scotland and Wales) published a revised version of the Scrapie Advisory Notes for Farmers in October 2009 (Anon, 2009). The Scrapie Advisory Notes are intended to assist sheep or goat keepers in fulfilling their legal obligations to immediately report to the veterinary authorities any animal that they suspect of being affected with scrapie. The notes are not designed to be a complete guide to a sheep or goat keeper’s obligations but a set of guidelines to encourage farmers to be vigilant and report suspect cases. Despite these efforts, scrapie has a low profile in the farming community and passive surveillance is no longer a source of new cases. With regards to the flock prevalence, the estimate calculated using an extension of the back-calculation model is considerably lower than previous estimates. However the different nature of the methods applied makes any comparison problematic. Capture–recapture methods rely on the comparison of lists of cases obtained from independent detection protocols (Brittain and Böhning, 2009), although the study of single case lists have also been applied, to estimate disease prevalence. For example, the first CRC model applied to three surveillance sources, one of which was SND (i.e. passive surveillance) and using 2002–2003 data reported an estimated number of infected holdings of 1653 (95% CI: 354–6438). The estimate reported in this study reflects a 90% reduction in the 10-year period, acknowledging the different approaches followed to produce such results, which is consistent with the large reduction in animal prevalence (Fig. 2). The estimate of 215 infected flocks is also broadly consistent with the estimation reported by
the same authors (Del Rio Vilas and Böhning, 2008) of 393 infected flocks (95% CI: 213–573) using a different data source: the CSFS 2005–2006, although the estimate from the present study would suggest only a 45% reduction from the 2005–2006 estimate of holding prevalence, compared to an estimated 90% reduction in prevalence at the animal level in that time period. Although small in number, the removal of heavily-infected flocks in the early years of the application of the compulsory control measures may have left a number of flocks with low prevalence of infection subdued by the increase in the overall genetic resistance at flock level. The holding prevalence estimate of 215 implies a relatively high average number of infected sheep per infected holding (5737 infected sheep in 215 holdings, indicates 26.7 infected sheep per holding). This is highly influenced by the degree of over-dispersion estimated from the CSFS data. This over-dispersion indicates a high degree of variability in the within-flock prevalence, and suggests the possibility of a few flocks having potentially large numbers of infected sheep. High prevalence flocks have been observed in GB, such as an infected large flock in which more than 200 cases were confirmed in 2011–2012 (Ortiz-Pelaez and Arnold, unpublished results). In one longitudinal study of 15 individual commercial infected sheep flocks, the flock-level prevalence estimates varied from 0 to 15.4% when culled animals were screened by immunohistochemistry for evidence of infection (Tongue et al., 2005). McIntyre et al. (2008) identified 415 cases of scrapie in 30 infected flocks, with numbers of cases per flock varying from 1 to 131, with seven flocks having only a single case of scrapie. The overall mean number of infected sheep per infected flock is further influenced by the greater likelihood that large flocks will be infected, with a significant influence of flock size on the probability of being infected (Green et al., 2007). If the implementation of the CSFS has influenced the within-flock prevalence distribution by removal of the high prevalence flocks, so that there is no longer such over-dispersion of the within-flock prevalence, this would have a significant effect on the estimate of holding-level prevalence. Repeating the estimation of holding prevalence but assuming a Poisson distribution in place of a negative binomial (and given the same mean within-flock prevalence of 2.2% apparent prevalence uprated by 2.2 for undetected infections) results in a holding prevalence of 404 flocks, almost double that assuming negative binomial within-flock prevalence. The estimate of flock prevalence also did not take into account the distribution of genotypes at the flock level i.e. there will be some flocks with a high proportion of scrapie resistant animals that are likely, if infected, to have a very low prevalence of within-flock infection, and other flocks with many susceptible animals with a high within flock prevalence. To incorporate the distribution of genotypes at the flock level, an analysis would need to account for the distribution of different flock genotype profiles in the national flock, for which there is limited data. Instead, the current approach relies on the estimate of within-flock prevalence from the cull of 121 CSFS flocks during the period 2005–2007, approximately half of all culled scrapieaffected flocks in GB.
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
The estimates of flock and animal prevalence confirm the decline of the burden of disease in GB and the combined effectiveness of the measures applied. It is difficult to apportion the decline in prevalence between each of the major control measures, the CSFS and the NSP. However, some insight may be gained from earlier predictions of the impact of breeding programmes on the animal prevalence in GB (Gubbins and Roden, 2006). Under the scenario where the proportion of sheep in NSP groups IV and V roughly halved (the scenario most similar to the observed decline in GB between 2002 and 2013, a drop from 12% to 7% (Ortiz-Pelaez et al., 2014; Appendix E)) Gubbins and Roden (2006) predicted an approximate halving of the prevalence of scrapie in GB. Assuming the modelling results from Gubbins and Roden (2006) are valid, this would suggest that the CSFS had a larger impact than the NSP on reducing sheep scrapie in GB, but that the NSP still had a significant effect. The estimates of animal and flock prevalence, while showing a large decline since 2005, also suggest that there are still a number of undetected flocks in GB with at least one infected animal. The delay in the detection due to the diminishing surveillance efforts, the relaxation of the control measures and the uncertainties on the current level of genetic resistance derived from the end of national schemes may all favour within-flock transmission leading to a higher prevalence in the remaining affected flocks than would have been expected at this stage of the epidemic (after nearly 10 years of enforcing the control measures and 8 years of breeding for resistance) given the within-flock prevalence distribution from the CSFS flocks (Fig. 4). This was showed in a large outbreak detected incidentally in a large flock where a total of 122 cases were confirmed in 2011, making it one of the largest outbreaks of classical scrapie in a single holding ever detected in GB (Ortiz-Pelaez et al., 2012). Further studies on the ability of the current active surveillance streams to detect scrapie infected animals/flocks at the reported prevalence levels and on the cost associated to them would be advisable.
8. Conclusions The decline in detection of scrapie cases continued from 2007, and this present study shows the prevalence in 2012 has dropped to 10% of that in 2005 indicating the effectiveness of the control measures. It also shows a bias in the destination of infected animals, with the majority of infected animals being detected in the fallen stock stream, and an extremely low proportion of animals detected as clinical suspects; this is very important in terms of the design of surveillance schemes for classical scrapie.
Acknowledgements This study was funded by the Department for Environment, Food and Rural Affairs (Defra) under the project TS5980. The helpful comments of two anonymous referees is also gratefully acknowledged.
249
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.prevetmed.2014.07.015. References Anon, 2009. Scrapie—advisory notes for farmers, http://archive.defra. gov.uk/foodfarm/farmanimal/diseases/atoz/bse/documents/scrapeng.pdf (accessed 21.03.14). Arnold, M.E., Ryan, J.B.M., Konold, T., Simmons, M.M., Spencer, Y.I., Wear, A., Chaplin, M., Stack, M., Czub, S., Mueller, R., Webb, P.R., Davis, A., Spiropoulos, J., Holdaway, J., Hawkins, S.A.C., Austin, A.R., Well s, G.A.H., 2007. Estimating the temporal relationship between PrPSc detection and incubation period in experimental bovine spongiform encephalopathy of cattle. J. Gen. Virol. 88, 3198–3208. Böhning, D., Del Rio Vilas, V.J., 2008. Estimating the hidden number of scrapie affected holdings in Great Britain using a simple, truncated count model Allowing for Heterogeneity. J. Agric. Biol. Environ. Stat. 13, 1–22. Bozetta, E., Nappi, R., Crudeli, S., Meloni, D., Varello, K., Loprevite, D., Melis, P.G., Mazza, M., Colussi, S., Ingravalle, F., Ru, G., Nonno, R., Ligios, C., 2011. Comparative performance of three TSE rapid tests for surveillance in healthy sheep affected by scrapie. J. Virol. Methods 173, 161–168. Brittain, S., Böhning, D., 2009. Estimators in capture recapture with two sources. Adv. Stat. Anal. 93, 23–47. Del Rio Vilas, V.J., Sayers, R., Sivam, K., Pfeiffer, D., Guitian, J., Wilesmith, J.W., 2005. A case study of capture–recapture methodology using scrapie surveillance data in Great Britain. Prev. Vet. Med. 67, 303–317. Del Rio Vilas, V.J., Böhning, D., 2008. Application of one-list capture–recapture models to scrapie surveillance data in Great Britain. Prev. Vet. Med. 85, 253–266. Dexter, G., Tongue, S.C., Heasman, L., Bellworthy, S.J., Davis, A., Moore, S.J., Simmons, M.M., Sayers, A.R., Simmons, H.A., Matthews, D., 2009. The evaluation of exposure risks for natural transmission of scrapie within an infected flock. BMC Vet. Res. 5, 38. Green, D.M., Del Rio Vilas, V.J., Birch, C.P.D., Johnson, J., Kiss, I.Z., McCarthy, N.D., Kao, R.R., 2007. Demographic risk factors for classical and atypical scrapie in Great Britain. J. Gen. Virol. 207, 3486–3492. Gubbins, S., Simmons, M.M., Sivam, K., Webb, C.R., Hoinville, L.J., 2003. Prevalence of scrapie infection in Great Britain: interpreting the results of the 1997–1998 abattoir survey. Proc. R. Soc. B. 270, 1919–1924. Gubbins, S., Roden, J.A., 2006. Breeding programmes for TSE resistance in British sheep II. Assessing the impact on the prevalence and incidence of scrapie. Prev. Vet. Med. 73, 17–31. Gubbins, S., 2008. Prevalence of sheep infected with classical scrapie in Great Britain: integrating multiple sources of surveillance data for 2002. J. R. Soc. Interface 5, 1343–1351. Gubbins, S., McIntyre, M., 2009. Prevalence of sheep infected with classical scrapie in Great Britain, 1993–2007. Epidemiol. Inf. 137, 787–791. Healy, A.M., Weavers, E., McElroy, M., Gomez-Parada, M., Collins, J.D., O’Doherty, E., Sweeney, T., Doherty, M.L., 2003. The clinical neurology of scrapie in Irish sheep. J. Vet. Intern. Med. 17, 908–916. Hoinville, L., McLean, A.R., Hoek, A., Gravenor, M.B., Wilesmith, J., 1999. Scrapie occurrence in Great Britain. Vet. Rec. 145, 405–406. Konold, T., Bone, G., Ortiz-Pelaez, A., Tortosa, R., Clifford, D., Dexter, G., Simmons, M.M., Spiropoulos, J., Berthelin-Baker, C.F., 2009. Associations of clinical signs and prion protein genotypes in British sheep with scrapie. Dtsch. Tierärztl. Wochenschr. 116, 380–388. Matthews, L., Coen, P.G., Foster, J.D., Hunter, N., Woolhouse, M.E.J., 2001. Population dynamics of a scrapie outbreak. Arch. Virol. 146, 1173–1186. McIntyre, K.M., Gubbins, S., Goldmann, W., Hunter, N., Baylis, M., 2008. Epidemiological characteristics of classical scrapie outbreaks in 30 sheep flocks in the United Kingdom. PLoS ONE 3, e3994. Ortiz-Pelaez, Á., Del Rio Vilas, V., 2009. Within-holding prevalence of sheep classical scrapie in Great Britain. BMC Vet. Res. 5, 1. Ortiz-Pelaez, A., Arnold, M.E., Warner, R., 2012. Sheep and goats scrapie surveillance 2011. Joint Descriptive Report for Great Britain, 2012, http://www.defra.gov.uk/ahvla-en/files/pub-tse-stats-scrapie.pdf
250
M. Arnold, A. Ortiz-Pelaez / Preventive Veterinary Medicine 117 (2014) 242–250
Ortiz-Pelaez, A., Thompson, C.E., Dawson, M., 2014. The impact of the National Scrapie Plan on the PRNP genotype distribution of the British national flock, 2002–2012. Vet. Rec., http://dx.doi.org/ 10.1136/vr.102255. Philipp, W.J., Kollmorgen, N., van Iwaarden, P., Charout-Got, J., Contreras Lopez, M., Goll, M., LeGuern, L., Schimmel, H., Vodrazka, P., 2005. The evaluation of rapid post mortem tests for the diagnosis of TSE in sheep. In: Report by the Institute for Reference Materials and Methods 2005. European Commission, See http://irmm.jrc.ec.europa.eu/ activities/TSE testing/Documents/040616 scrapie report global.pdf (accessed 21.03.14). Sivam, S.K., Baylis, M., Gravenor, M.B., Gubbins, S., 2006. Descriptive analysis of the results of an anonymous postal survey of the occurrence of
scrapie in Great Britain in 2002. Vet. Rec. 158, 501–506. Tongue, S.C., Webb, P., Simmons, M.M., Gubbins, S., 2005. Prevalence of scrapie infection in cull animals from 14 scrapie-affected flocks in Great Britain. Vet. Rec. 157, 480–482. Tongue, S.C., Pfeiffer, D.U., Warner, R., Elliott, H., Del Rio Vilas, V., 2006. Estimation of the relative risk of developing clinical scrapie: the role of prion protein (PrP) genotype and selection bias. Vet. Rec. 158, 43–50. Wells, G.A.H., Konold, T., Arnold, M.E., Austin, A.R., Hawkins, S.A.C., Stack, M., Simmons, M.M., Lee, Y.H., Gavier-Widén, D., Dawson, M., Wilesmith, J.W., 2007. Bovine spongiform encephalopathy: the effect of oral exposure dose on attack rate and incubation period in cattle. J. Gen. Virol. 88, 1363–1373.
