| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 UMR Interactions Hôtes Agents Pathogènes, Ecole Nationale Vétérinaire de Toulouse, 23 chemin des Capelles, 31076 Toulouse Cedex, France
2 EMI 0338 (Biostatistique), Institut de Santé Publique et Développement, Université Victor Segalen Bordeaux 2, 146 rue Léo Saignat, 33076 Bordeaux Cedex, France
3 Station d'Amélioration des Animaux, Institut National de la Recherche Agronomique, BP 27, 31326 Castanet-Tolosan cedex, France
4 Centre Départemental d'Elevage Ovin, Quartier Ahetzia, 64130 Ordiarp, France
5 Direction Départementale des Services Vétérinaires, Cours Lyautey, 64000 Pau, France
Correspondence
Fabien Corbière
fabien.corbiere{at}isped.u-bordeaux2.fr
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
). Under natural conditions, it is considered that scrapie infection in sheep occurs mainly after an early oral contamination (around birth) (Andréoletti et al., 2000; Heggebo et al., 2000). Scrapie clinical onset generally occurs between 2 and 7 years (Detwiler & Baylis, 2003).
In prion diseases, the accumulation of an abnormal isoform (PrPSc) of a normal cellular protein (PrPC) in tissues from infected individuals is currently considered as a disease hallmark. Most of the diagnostic tests currently available are based on biochemical detection of the abnormal protein (McKinley et al., 1983; Race et al., 2001). However, post-mortem tests, as carried out in the current European surveillance programme (rapid test on the obex), are reliable only for detection of infected animals in the second half of the incubation period.
Because of long TSE incubation periods, data analysis is difficult without reference to flock demography and management. Indeed, infected individuals could be culled or die from other causes before clinical onset (intercurrent diseases, economic reasons). In this situation, no reliable information will be available about their infectious status (Begara-McGorum et al., 2000; Ryder et al., 2001; Thorgeirsdottir et al., 2002; Billinis et al., 2004).
Evaluation of genetic and environmental risk factors in scrapie has been conducted mainly using case–control designs, in which a set of affected animals is compared with their healthy flock-mates or to a reference population (Hunter et al., 1997; Thorgeirsdottir et al., 1999; Tranulis et al., 1999; O'Doherty et al., 2002; Acin et al., 2004; Baylis et al., 2004; Billinis et al., 2004; Tongue et al., 2006). Such approaches have revealed that TSE susceptibility in sheep is controlled mainly by polymorphisms at codons 136 (T, V, A), 154 (R, H) and 171 (R, Q, H, K) of the PRNP gene (Clouscard et al., 1995; Hunter et al., 1996). V136R154Q171/VRQ, ARQ/VRQ and ARQ/ARQ animals are usually considered the most susceptible to scrapie, whereas homozygous or heterozygous AHQ and heterozygous ARR animals show only marginal susceptibility, ARR/ARR sheep being considered to be fully clinically resistant (Detwiler & Baylis, 2003).
Surveys based on long-term individual monitoring of an exposed population are less subject to sampling bias (Tongue et al., 2006). Consequently, they should be considered as more relevant than case–control or cross-sectional studies for an accurate evaluation of the effect of environmental or genetic factors on infection rate and incubation period.
‘Cure models' are part of the mixture models family (Bohning & Seidel, 2003). In ‘mixture cure models’, it is considered that the studied population is a mixture of susceptible (i.e. that may undergo the event of interest) and non-susceptible individuals (i.e. that will never undergo the event of interest) (Farewell, 1982). Unlike classical survival analysis, they allow a separate estimation of covariate effects on incidence and incubation length. Cure models also allow estimation of the proportion of healthy (or conversely infected) individuals in a population, including individuals that did not last the total length of the study (Lam et al., 2005).
In this study, we propose a model based on a ‘mixture cure models' approach for scrapie epidemiological analysis. Robustness and reliability boundaries of the model were assessed by simulations before analysing data collected over 6–9 years in six naturally scrapie-affected flocks in Pyrénées Atlantiques, France.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
The model assumes that: (i) most or all deaths from scrapie occur during a determined age-period, (ii) monitoring is long enough for clinical onset to have appeared in most of the infected animals and (iii) the longer an animal lives, the lower the probability of it being scrapie infected. Animals which live longer than the last observed scrapie clinical case are considered to have an extremely low probability (if not zero) of incubating scrapie.
If U is the indicator denoting SI animals (i.e. U=1 if the animal is scrapie-infected and U=0 if non-infected) and T is a non-negative random variable denoting the failure time of interest, defined only if U=1, the mixture cure model is given as follows:
|
|
(z)=P(U=1|z) is the probability of being infected given a covariate vector z=(z1,..,zq)'. S(t|U=1,x)=P(T >t|U=1,x) is the conditional survival function for SI animals given a covariate vector x=(x1,..,xm)' (it may include the same covariate as z). The use of conditional attached to this function is to stress that the distribution of time refers not to the whole group of animals but only to the animals that are in the SI group. Note that S(t|x,z) 
1–(z) as t
, where 1–(z) represents the proportion of non-infected animals. When =1, that is when no SF portion is assumed, the model reduces to the traditional survival model. Whether the inclusion of a proportion of SF animals leads to a significantly better fit to the data than a traditional survival model with no SF animals can be tested by the deviance test statistics proposed by Maller & Zhou (1996).
Various parametric and semiparametric approaches have been proposed for mixture cure models (Peng & Carriere, 2002; Lam et al., 2005). For modelling the influence of exploratory variables on the incidence, a logistic regression model is usually chosen (Kuk & Chen, 1992; Peng & Dear, 2000):
|
|
is the vector of regression parameters associated with z and contains an intercept. The conditional survival function of infected animals is modelled through the semiparametric Cox proportional hazards (PH) model (Cox, 1972), which is given by |
|
is the vector of regression parameters associated with x and S0(t|U=1) is the baseline conditional survival function, which is left unspecified.
Through the vectors of regression parameters and , the mixture survival model is able to separate the effects of the covariates on incidence and latency. An estimate of the true proportion of SI animals, SIpop, given z and x, is provided by taking the mean of the individual probabilities Pi(U=1|zi,xi):
|
|
i is the censoring indicator with 1 if ti is uncensored and 0 otherwise. Obviously, if i=1, then Pi(U=1)=1. When i=0, then Pi(U=1|zi, xi) will depend on the survival length and will drop to zero as t. Note that the better the model fits the data, through covariate vectors z' and x', the more accurate is the estimation of the proportion of SI animals.
Simulation studies.
Simulations were conducted to test (i) the ability of the model to estimate the proportion of SI animals and to discriminate covariate effects on the infection risk and incubation duration and (ii) the effect of individual monitoring length on model outputs. We assumed that non-infected (SF) animals would never die from scrapie. Consequently, observations conducted on SF animals were right-censored. SI animals either died from clinical scrapie (uncensored records) or were eliminated from the flock before clinical onset (right-censored). Simulations were performed using (i) genetic and biological parameters (infection rates, ages at clinical death and flock demography) described in scrapie outbreaks or already used in mathematical modelling (simulation design 1) (Matthews et al., 2001; Hagenaars et al., 2003; Hopp et al., 2003; Baylis et al., 2004; Eglin et al., 2005; Touzeau et al., 2005) and (ii) ages at death from scrapie reflecting observations made in our dataset (simulation design 2).
The capacity of the model to separate covariate effects on incidence and incubation length (age at death) was assessed by generating two independent binary covariates, one (Z1) affecting only the incidence and the other (Z2) affecting only the incubation duration. The incidence is given the logistic form
|
|
1 and 2 are the effects of covariates Z1 and Z2 on the proportion of infected individuals, respectively. Since Z2 should have no effect on the incidence, 2 was set to 0. Thus, the proportion of infected animals is |
|
|
|
0=–0.5 and 1=–1, so that the corresponding proportions of infected sheep are 37.7 % (animals with Z1=0) and 18.22 % (animals with Z1=1), respectively.
The log-normal distribution was used as the distribution function for life expectancies of infected animals, with survival function
|
|
is the distribution function of the standard normal law. In contrast to the incidence portion, 1 was set to 0, because Z1 should have no effect on latency, whatever the value of Z2. In simulation design 1, the scale (µ) and shape (
) parameters for the log-normal distribution function were set to 1.2 and 0.4, respectively. In the absence of censoring, events of interest (scrapie deaths) were allowed to occur at median age 3.3 years (interquartile range 2.5–4.2,) for individuals with Z2=0. We set 2=0.3 so that infected individuals with Z2=1 would die later, at median age 4.4 years (interquartile range 3.4–5.7). In simulation design 2, we set µ=ln 2, =0.5, 1=0 and 2=0.375, so that median ages at clinical onset were 2.00 (interquartile range 1.5–2.8) and 3.00 years (interquartile range 2.1–4.2) (Fig. 1a).
|
,
), with shape () and scale () parameters. Thirteen different scenarios were investigated with median life expectancy (meaning monitoring length) ranging from 2.5 to 9.5 years. These scenarios covered a large panel of flock management policies and demography. For each scenario and simulation design, 500 independent datasets, each consisting of 500 individuals, were generated and submitted to model analysis. The absolute biases [B(
)=
i(i–c0)/500] and mean squared errors [MSE()=i(i–c0)2/500], where i are the estimates of c0, were computed for the five parameter estimates. Sample generation and model computations were performed using SAS software (SAS-PC system, Version 8.2 for Windows; SAS Institute).
Flocks.
Investigations were carried out on six naturally scrapie-affected dairy flocks, bred by private farmers, in Pyrénées Atlantiques, France. These flocks had been involved in a long-term scrapie epidemiology research project since 1994 (flock C) and 1998 (the other five flocks). Sheep were all Manech red-faced pure-breeds. Table 1 shows, for each flock, the mean flock size and the estimated year of first occurrence of scrapie. The high incidences of scrapie clinical suspicions (confirmed or not) in ewes born prior to enrolment in the research project suggest different, but nonetheless high, infection pressures (Table 1).
|
To comply with the requirement of adequate monitoring length and the provision of high-quality data (including reliable diagnosis), only a few birth cohorts were considered for the analysis within each flock. Birth cohorts were selected in which all scrapie clinical suspicions were confirmed by histopathology and complete PrP genotype profiles were available.
The dataset submitted for analysis consisted of the 1998 birth cohort (born between October and December 1997) for flock A, the 1999 birth cohort (born between October and December 1998) for flocks B, D, E and F and the 1995, 1996 and 1997 birth cohorts in flock C (animals born in November and December 1994, 1995 and 1996, respectively) (Table 2). Only homebred animals were included in the analysis, while purchased sheep (n=10) were not considered. In total, our sample comprised 641 sheep, including 170 scrapie clinical cases.
|
). All the sheep included in our initial sample were not examined, either because of failure to track these animals during the elimination process or because they were still alive at the time of writing (January 2006). From these latter animals, palatine tonsils biopsies were performed each year from 2002 to 2005 (inclusive) for PrPSc immunohistochemistry detection (Andréoletti et al., 2002). To account for missing information, a one-sided 95 % confidence interval for the true proportion of infected animals in each PrP genotype group was calculated using the hypergeometric law.
In the mixture cure model analysis, PrP genotype and the flock were used as covariates and age at death from clinical scrapie was considered as the survival measurement. Ninety-five per cent confidence intervals for adjusted odds ratios (OR) from the logistic part and adjusted relative risks (RR) from the Cox PH part of the mixture cure model were computed using the bias corrected, accelerated bootstrap method (Davison & Hinkley, 1997).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
). Biases and MSE were acceptable for inference when the monitoring time was longer than the median (theoretical) incubation duration. For the first set of simulated incubation times (median age at death 3.3 and 4.45 years), the estimates of the proportion of infected animals (0) and the effects of Z2 on the incidence (2) were highly biased (absolute bias over 0.1) for median monitoring times of less than 5.5 years (Fig. 1b). For the second set of simulated incubation durations (median age at death 2 and 3.3 years), similar results were obtained for median monitoring times under 3.5 years (Fig. 1c). In our population, median age at death in scrapie-affected animals was 1.7 years (Fig. 1a; Table 3) and the median monitoring time was 4.90 years (Table 4). Under these conditions, the model outputs were expected to be relevant in the analysis of our dataset.
|
|
). The genetic structure of the sample is consistent with previously reported PrP polymorphism frequencies in the Manech red-faced breed (Palhiere et al., 2002).
Because of small numbers of individuals in some PrP genotypes, animals were grouped according to their level of susceptibility to classical scrapie (DEFRA, 2003). ARQ/ARQ, AHQ/AHQ and AHQ/ARQ sheep were considered in a medium-risk group (S/S; n=343). ARR/ARQ, ARR/AHQ and ARR/VRQ animals were included in a low-risk group (R/S; n=212). ARQ/VRQ, AHQ/VRQ and VRQ/VRQ animals were included in a single high-risk group (VRQ/x; n=49) (Table 3). Considering these PrP genotype groups, the genetic structure was not statistically different in the six selected flocks (chi-square test with 15 degrees of freedom: 
2=15.95, P=0.38).
Scrapie clinical cases
Clinical scrapie cases were mainly observed in ARQ/ARQ (n=131; 77.06 %) and ARQ/VRQ (n=28; 16.47 %) genotypes, while heterozygote ARR were poorly affected (R/S sheep: n=9; 5.29 %) (Tables 2 and 3). High incidences in susceptible PrP genotypes ARQ/ARQ and ARQ/VRQ suggested a high infection pressure. No clinical cases were observed in ARR/ARR (n=37), ARR/VRQ (n=3), ARR/ARH (n=3) or AHQ/VRQ (n=3) animals. However, the number of animals with these last three genotypes was too small to draw any conclusions.
Kaplan–Meier plots of the survival distribution functions for scrapie clinical occurrence indicate the absence of new scrapie cases after 5.54 years, whatever the genotype group considered (Fig. 2). This lack of new clinical cases fulfils two basic requirements that allow the application of the Cox PH mixture cure model, i.e. (i) most or all deaths due to scrapie occurred in a defined age period and (ii) the monitoring length was sufficient to allow almost all infected animals to show clinical signs.
|
). However, 94.24 % of these symptomless sheep lived longer than the median age at clinical onset (1.70 years), and 72.12 % reached 3.57 years old, which represented the 90th percentile of the age distribution of scrapie clinical cases. Because of the implementation of French TSE legislation at the beginning of 2003, breeders had to remove VRQ carrier animals from scrapie-affected flocks. Consequently, VRQ animals, mainly ARR/VRQ, were eliminated earlier than expected and had a statistically shorter follow-up than R/S animals (analysis of variance; F431,2=13.13, P<10–4).
Active detection of subclinically infected sheep
Of the 471 clinically healthy sheep (with no clinical scrapie) eliminated during the study, 220 (46.71 %) were submitted to post-mortem for PrPSc detection (mean age 4.72 years; 95 % confidence interval 1.87–7.57 years). Sampled animals represented 70 % of VRQ/x and 72.04 % of S/S but only 24.13 % of R/S and 13.51 % of R/R (Table 5).
|
).
Results from the mixture cure model analysis
The deviance statistic test (201=30.93; P<10–4) indicated that incorporating a scrapie-free fraction provided a better fit to the data than the traditional Cox PH model and that the estimated effects were more relevant.
Proportion of infected animals
ARR/ARR animals were not included in the analysis, because there were no confirmed clinical cases of this genotype. According to the proposed model, the predicted number of subclinically infected animals was 20.75 (respectively 3.65 in the R/S group, 14.83 in the S/S group and 3.22 in the VRQ carrier group; point estimate minus number of clinical cases) (Table 5). Strikingly, these predicted numbers were in close agreement with those obtained by the active detection of subclinical cases.
Genotype influence on incidence and incubation duration
Results from the Cox mixture cure model indicated that ARR heterozygote animals were at lower risk of infection than S/S animals (Table 6). Conversely, VRQ allele carriers (excluding ARR/VRQ) were at higher risk of being infected. Age at clinical onset was significantly greater in R/S animals when compared with S/S animals. No significant difference was found between S/S and VRQ allele carriers (Table 6).
|
). The risk for ARQ/ARQ animals of being infected was significantly lower in flocks A, D and E than in flock C (Table 6). No significant difference was observed between flocks C, B and F. Interestingly, age at clinical onset was significantly less in some flocks with a lower infection risk (flock D compared with flock C). Meanwhile, age at clinical onset could be significantly different in flocks harbouring a similar infection risk: infected animals from flock F had delayed clinical onset compared with flock C while, in flock B, it was shortened. Taken together, these results strongly suggest that age at clinical onset and infection risk are not associated parameters. |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
; Diaz et al., 2005). Using survival analysis, they aimed to determine the influence of PrP genotype, rearing type and dam clinical status on individual risk of developing clinical scrapie. Survival analysis assumes intrinsically that if, in an exposed population, a long enough and complete surveillance of individuals is possible, each would experience the event of interest, i.e. clinical scrapie. For TSE, this hypothesis is obviously inadequate. Indeed, under natural exposure, ARR/ARR animals and a large proportion of susceptible-genotype animals will remain uninfected (Elsen et al., 1999).
Because a mixed population of susceptible and non-susceptible individuals is considered, ‘mixture cure models' appear to be an attractive approach for scrapie epidemiological analysis. However, several conditions must be fulfilled for their sound application. Such constraints require hypotheses about scrapie pathogenesis and biology. Amongst the basic hypotheses we considered were that animals born in an infected flock, if not infected in their early life, would remain negative. Under natural exposure conditions, scrapie contamination is considered to occur preferentially around birth (Andréoletti et al., 2000; Heggebo et al., 2000; van Keulen et al., 2000). We therefore hypothesized that age at death from scrapie (clinical stage) was a relevant approximation of the incubation period. Moreover, animal susceptibility seems to decrease dramatically with age (Hourrigan et al., 1979; Andréoletti et al., 2000). Clinical cases have been reported in young and adult susceptible animals introduced to infected flocks (Hourrigan et al., 1979; Ryder et al., 2004) but the importance, relative to neonatal contamination, of such lateral transmission in adult sheep could not be estimated. The other main hypothesis we made was that very few (if not zero) infected individuals would be alive at the end of the study. The observed survival distribution plots were consistent with this hypothesis. However, existence of long-term subclinical carriers remains a major question of scrapie epidemiology. Currently, it is impossible to assume that an apparently healthy animal (whatever the test used to establish infectious status) is not incubating scrapie. Recent description of atypical cases or Nor98 cases in old and apparently healthy animals, and difficulties in assessing the diagnosis, sustain the ‘long-term subclinical carriers' hypothesis (Benestad et al., 2003; Le Dur et al., 2005). However, atypical scrapie occurs at a very low detected prevalence level (3–11 cases per 10 000 examined) and, in most cases, only one to three cases could be detected in stamped-out affected flocks (De Bosschere et al., 2004; Onnasch et al., 2004; Orge et al., 2004). This implies that approximately 0.2–0.8 sheep could have been infected with atypical scrapie in the considered flocks, which is negligible when compared with the number of classical scrapie cases in the studied cohorts (n=170). At the population level, the influence of an atypical case on the model outputs was considered to be negligible.
Finally, even if the hypothesis of some adult lateral transmission and long-term subclinical carriers could not definitely be ruled out, both phenomena seemed marginal enough in our population to avoid major transgression from application of the model. From simulations, major biases were observed only when the monitoring length was shorter than the median (theoretical) incubation duration. Similar trends were obtained by Yu et al. (2004) when studying the influence of the follow-up length on the cure fraction estimation for several human cancers. The monitoring length in the studied sheep was long enough to ensure small biases for the estimates of PrP genotype and flock effects.
Asymptomatic culled animals
The mixture cure model approach enabled us to estimate the number of infected individuals and included those eliminated while incubating the disease. Model outputs and laboratory findings were in close agreement and indicated that a very small number of sheep were removed while incubating scrapie.
This result is consistent with observations from another study based on a longitudinal survey in a Texel flock (Baylis et al., 2002). It contrasts, however, with other publications in which large numbers of scrapie-incubating animals were reported (Thorgeirsdottir et al., 2002; Billinis et al., 2004). Similarly, the modelling of a scrapie outbreak in a Cheviot flock predicted a high ratio of infections to cases (2.2 : 1) (Matthews et al., 2001).
Discrepancies between these results certainly lay in the data collection plan. Studies reporting a large proportion of asymptomatic animals were based on cross-sectional designs, with data collected at stamping-out. In our study, most sheep were culled after a long individual monitoring period which allowed scrapie clinical onset. As indicated by our simulations, shorter monitoring lengths, as modelled by Matthews et al. (2001) (median length 3.00 years), would have resulted in the observation of fewer clinical cases and a larger number of subclinical cases.
Genetic susceptibility to scrapie and incubation period
Comparison of the fit provided by the mixture cure model and the traditional Cox PH model indicated that our approach was more relevant when analysing PrP genotype and flock effects. According to our results, with ARQ homozygote animals as the baseline, VRQ carriers were at higher risk of infection and ARR heterozygotes at lower risk. This is consistent with most published studies based on data collected from culled flocks (Thorgeirsdottir et al., 1999; Tranulis et al., 1999; Acin et al., 2004; Billinis et al., 2004).
Incubation length is a major feature of TSE phenotype. In our population, while clinical signs were delayed in ARR heterozygotes compared with ARQ homozygotes, no difference could be observed between ARQ/VRQ carriers and ARQ homozygotes. A similar phenomenon was observed in an Irish flock (O'Doherty et al., 2002). However, it differed from estimations obtained in a French Romanov flock (Elsen et al., 1999) and in a Texel flock (Baylis et al., 2002). In both these naturally affected scrapie flocks, significant differences in age at death were reported between ARQ/ARQ and ARQ/VRQ.
In sheep, experimental challenge has indicated that incubation period depends on both sheep genotype and TSE isolate. While most scrapie isolates will produce shorter incubation periods in VRQ allele carriers, other TSE agents such as BSE behave differently (Foster et al., 2001; Jeffrey et al., 2006). In this context, difference in agent (strain) could be a possible explanation for the observed variability.
In rodent scrapie models, it has been demonstrated that variations in attack rate and incubation length can be observed according to the infectious dose. Low infectious dose could lead to lengthening of the incubation period and decreased infection efficiency (Kimberlin & Wilesmith, 1994; Jacquemot et al., 2005). In natural scrapie, there is no available estimation of the actual infectious pressure. Because of differences in the observed incidences, infection pressure is usually considered to be different according to the cohort considered within a flock or between flocks (Baylis et al., 2002; Touzeau et al., 2005). In our study, duration of scrapie incubation appeared not to be associated with the infection rate. Age at clinical onset in ARQ/ARQ infected animals also clearly differed (shorter incubation period in our study) from values reported previously in animals bearing the same genotype (Woolhouse et al., 1998; Elsen et al., 1999; O'Doherty et al., 2002; Redman et al., 2002; Baylis et al., 2004). Taken together, these observations could suggest that biologically different scrapie agents are involved in the different flocks or that other factors linked to flock management could influence incidence and incubation.
To take these studies further, evaluation of agent biodiversity in the studied cohorts from these flocks is ongoing through biochemical studies and bioassays. Meanwhile, the effect of increasing infectious dose on incubation length in animals bearing similar PrP genotypes and which were orally contaminated at birth is under investigation.
The mixture cure model presented here has provided an interesting tool to analyse data collected from longitudinal surveys in naturally affected scrapie flocks. Its main constraint is the requirement for a sufficiently long individual monitoring period. Finally, because such models allow for the combinatory analysis of several covariate effects, they should be considered as a potentially powerful tool for epidemiological analysis in animal diseases.
|
ACKNOWLEDGEMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Andréoletti, O., Berthon, P., Marc, D., Sarradin, P., Grosclaude, J., van Keulen, L., Schelcher, F., Elsen, J. M. & Lantier, F. (2000). Early accumulation of PrPSc in gut-associated lymphoid and nervous tissues of susceptible sheep from a Romanov flock with natural scrapie. J Gen Virol 81, 3115–3126.
Andréoletti, O., Berthon, P., Levavasseur, E., Marc, D., Lantier, F., Monks, E., Elsen, J. M. & Schelcher, F. (2002). Phenotyping of protein-prion (PrPsc)-accumulating cells in lymphoid and neural tissues of naturally scrapie-affected sheep by double-labeling immunohistochemistry. J Histochem Cytochem 50, 1357–1370.
Baylis, M., Goldmann, W., Houston, F., Cairns, D., Chong, A., Ross, A., Smith, A., Hunter, N. & McLean, A. R. (2002). Scrapie epidemic in a fully PrP-genotyped sheep flock. J Gen Virol 83, 2907–2914.
Baylis, M., Chihota, C., Stevenson, E., Goldmann, W., Smith, A., Sivam, K., Tongue, S. & Gravenor, M. B. (2004). Risk of scrapie in British sheep of different prion protein genotype. J Gen Virol 85, 2735–2740.
Begara-McGorum, I., Clark, A. M., Martin, S. & Jeffrey, M. (2000). Prevalence of vacuolar lesions consistent with scrapie in the brains of healthy cull sheep of the Shetland Islands. Vet Rec 147, 439–441.
Benestad, S. L., Sarradin, P., Thu, B., Schonheit, J., Tranulis, M. A. & Bratberg, B. (2003). Cases of scrapie with unusual features in Norway and designation of a new type, Nor98. Vet Rec 153, 202–208.
Billinis, C., Psychas, V., Leontides, L., Spyrou, V., Argyroudis, S., Vlemmas, I., Leontides, S., Sklaviadis, T. & Papadopoulos, O. (2004). Prion protein gene polymorphisms in healthy and scrapie-affected sheep in Greece. J Gen Virol 85, 547–554.
Bohning, D. & Seidel, W. (2003). Editorial: recent developments in mixture models. Comput Stat Data Anal 41, 349–357.[CrossRef]
Clouscard, C., Beaudry, P., Elsen, J. M., Milan, D., Dussaucy, M., Bounneau, C., Schelcher, F., Chatelain, J., Launay, J. M. & Laplanche, J. L. (1995). Different allelic effects of the codons 136 and 171 of the prion protein gene in sheep with natural scrapie. J Gen Virol 76, 2097–2101.
Cox, D. R. (1972). Regression models and life-tables. J R Stat Soc Ser B Methodol 34, 187–220.
Davison, A. C. & Hinkley, D. V. (1997). Bootstrap Methods and their Application. Cambridge: Cambridge University Press.
De Bosschere, H., Roels, S., Benestad, S. L. & Vanopdenbosch, E. (2004). Scrapie case similar to Nor98 diagnosed in Belgium via active surveillance. Vet Rec 155, 707–708.
DEFRA (2003). National Scrapie Plan for Great Britain. London: Department for Food, Environment and Rural Affairs. http://www.defra.gov.uk/animalh/bse/othertses/scrapie/nsp/publicatsrpts/tables.htm
Detwiler, L. A. & Baylis, M. (2003). The epidemiology of scrapie. Rev Sci Tech 22, 121–143.[Medline]
Diaz, C., Vitezica, Z. G., Rupp, R., Andréoletti, O. & Elsen, J. M. (2005). Polygenic variation and transmission factors involved in the resistance/susceptibility to scrapie in a Romanov flock. J Gen Virol 86, 849–857.
Eglin, R. D., Warner, R., Gubbins, S., Sivam, S. K. & Dawson, M. (2005). Frequencies of PrP genotypes in 38 breeds of sheep sampled in the National Scrapie Plan for Great Britain. Vet Rec 156, 433–437.
Elsen, J. M., Amigues, Y., Schelcher, F., Ducrocq, V., Andréoletti, O., Eychenne, F., Khang, J. V., Poivey, J. P., Lantier, F. & Laplanche, J. L. (1999). Genetic susceptibility and transmission factors in scrapie: detailed analysis of an epidemic in a closed flock of Romanov. Arch Virol 144, 431–445.[CrossRef][Medline]
Farewell, V. T. (1982). The use of mixture models for the analysis of survival data with long-term survivors. Biometrics 38, 1041–1046.[CrossRef][Medline]
Foster, J. D., Parnham, D., Chong, A., Goldmann, W. & Hunter, N. (2001). Clinical signs, histopathology and genetics of experimental transmission of BSE and natural scrapie to sheep and goats. Vet Rec 148, 165–171.
Fraser, H. (1976). The pathology of a natural and experimental scrapie. Front Biol 44, 267–305.[Medline]
Hagenaars, T. J., Donnelly, C. A., Ferguson, N. M. & Anderson, R. M. (2003). Dynamics of a scrapie outbreak in a flock of Romanov sheep – estimation of transmission parameters. Epidemiol Infect 131, 1015–1022.[CrossRef][Medline]
Heggebo, R., Press, C. M., Gunnes, G., Lie, K. I., Tranulis, M. A., Ulvund, M., Groschup, M. H. & Landsverk, T. (2000). Distribution of prion protein in the ileal Peyer's patch of scrapie-free lambs and lambs naturally and experimentally exposed to the scrapie agent. J Gen Virol 81, 2327–2337.
Hopp, P., Webb, C. R. & Jarp, J. (2003). Monte Carlo simulation of surveillance strategies for scrapie in Norwegian sheep. Prev Vet Med 61, 103–125.[CrossRef][Medline]
Hourrigan, J., Klingsporn, A. L., Clark, W. W. & de Camp, M. (1979). Epidemiology of scrapie in the United States. In Slow Transmissible Diseases of the Nervous System, pp. 331–356. Edited by S. B. Prusiner & W. J. Hadlow. New York: Academic Press.
Hunter, N., Foster, J. D., Goldmann, W., Stear, M. J., Hope, J. & Bostock, C. (1996). Natural scrapie in a closed flock of Cheviot sheep occurs only in specific PrP genotypes. Arch Virol 141, 809–824.[CrossRef][Medline]
Hunter, N., Goldmann, W., Foster, J. D., Cairns, D. & Smith, G. (1997). Natural scrapie and PrP genotype: case-control studies in British sheep. Vet Rec 141, 137–140.
Jacquemot, C., Cuche, C., Dormont, D. & Lazarini, F. (2005). High incidence of scrapie induced by repeated injections of subinfectious prion doses. J Virol 79, 8904–8908.
Jeffrey, M., Gonzalez, L., Chong, A., Foster, J., Goldmann, W., Hunter, N. & Martin, S. (2006). Ovine infection with the agents of scrapie (CH1641 isolate) and bovine spongiform encephalopathy: immunochemical similarities can be resolved by immunohistochemistry. J Comp Pathol 134, 17–29.[CrossRef][Medline]
Kimberlin, R. H. & Wilesmith, J. W. (1994). Bovine spongiform encephalopathy. Epidemiology, low dose exposure and risks. Ann N Y Acad Sci 724, 210–220.[Abstract]
Kuk, A. Y. C. & Chen, C.-H. (1992). A mixture model combining logistic regression with proportional hazards regression. Biometrika 79, 531–541.
Lam, K. F., Fong, D. Y. & Tang, O. Y. (2005). Estimating the proportion of cured patients in a censored sample. Stat Med 24, 1865–1879.[CrossRef][Medline]
Le Dur, A., Beringue, V., Andréoletti, O., Reine, F., Lai, T. L., Baron, T., Bratberg, B., Vilotte, J. L., Sarradin, P. & other authors (2005). A newly identified type of scrapie agent can naturally infect sheep with resistant PrP genotypes. Proc Natl Acad Sci U S A 102, 16031–16036.
Maller, R. A. & Zhou, X. (1996). Survival Analysis with Long-Term Survivors. Chichester: Wiley.
Matthews, L., Coen, P. G., Foster, J. D., Hunter, N. & Woolhouse, M. E. (2001). Population dynamics of a scrapie outbreak. Arch Virol 146, 1173–1186.[CrossRef][Medline]
McKinley, M. P., Bolton, D. C. & Prusiner, S. B. (1983). A protease-resistant protein is a structural component of the scrapie prion. Cell 35, 57–62.[CrossRef][Medline]
O'Doherty, E., Healy, A., Aherne, M., Hanrahan, J. P., Weavers, E., Doherty, M., Roche, J. F., Gunn, M. & Sweeney, T. (2002). Prion protein (PrP) gene polymorphisms associated with natural scrapie cases and their flock-mates in Ireland. Res Vet Sci 73, 243–250.[CrossRef][Medline]
Onnasch, H., Gunn, H. M., Bradshaw, B. J., Benestad, S. L. & Bassett, H. F. (2004). Two Irish cases of scrapie resembling Nor98. Vet Rec 155, 636–637.
Orge, L., Galo, A., Machado, C., Lima, C., Ochoa, C., Silva, J., Ramos, M. & Simas, J. P. (2004). Identification of putative atypical scrapie in sheep in Portugal. J Gen Virol 85, 3487–3491.
Palhiere, I., François, D., Elsen, J. M., Barillet, F., Amigues, Y., Perret, G. & Bouix, J. (2002). Allele frequencies of the PrP gene in 29 French sheep breeds. Possible use in selection for resistance to scrapie. In Proceedings of the Seventh World Congress on Genetics Applied to Livestock Production, Session 13, pp. 13–16. Montpellier, France.
Peng, Y. & Dear, K. B. (2000). A nonparametric mixture model for cure rate estimation. Biometrics 56, 237–243.[CrossRef][Medline]
Peng, Y. & Carriere, K. C. (2002). An empirical comparison of parametric and semiparametric cure models. Biom J 44, 1002–1014.[CrossRef]
Race, R., Raines, A., Raymond, G. J., Caughey, B. & Chesebro, B. (2001). Long-term subclinical carrier state precedes scrapie replication and adaptation in a resistant species: analogies to bovine spongiform encephalopathy and variant Creutzfeldt-Jakob disease in humans. J Virol 75, 10106–10112.
Redman, C. A., Coen, P. G., Matthews, L., Lewis, R. M., Dingwall, W. S., Foster, J. D., Chase-Topping, M. E., Hunter, N. & Woolhouse, M. E. (2002). Comparative epidemiology of scrapie outbreaks in individual sheep flocks. Epidemiol Infect 128, 513–521.[CrossRef][Medline]
Ryder, S. J., Spencer, Y. I., Bellerby, P. J. & March, S. A. (2001). Immunohistochemical detection of PrP in the medulla oblongata of sheep: the spectrum of staining in normal and scrapie-affected sheep. Vet Rec 148, 7–13.
Ryder, S., Dexter, G., Bellworthy, S. & Tongue, S. (2004). Demonstration of lateral transmission of scrapie between sheep kept under natural conditions using lymphoid tissue biopsy. Res Vet Sci 76, 211–217.[CrossRef][Medline]
Thorgeirsdottir, S., Sigurdarson, S., Thorisson, H. M., Georgsson, G. & Palsdottir, A. (1999). PrP gene polymorphism and natural scrapie in Icelandic sheep. J Gen Virol 80, 2527–2534.
Thorgeirsdottir, S., Georgsson, G., Reynisson, E., Sigurdarson, S. & Palsdottir, A. (2002). Search for healthy carriers of scrapie: an assessment of subclinical infection of sheep in an Icelandic scrapie flock by three diagnostic methods and correlation with PrP genotypes. Arch Virol 147, 709–722.[CrossRef][Medline]
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.
Touzeau, S., Chase-Topping, M. E., Matthews, L., Lajous, D., Eychenne, F., Hunter, N., Foster, J. D., Simm, G., Elsen, J. M. & Woolhouse, M. E. (2005). Modelling the spread of scrapie in a sheep flock: evidence for increased transmission during lambing seasons. Arch Virol 151, 735–751.
Tranulis, M. A., Osland, A., Bratberg, B. & Ulvund, M. J. (1999). Prion protein gene polymorphisms in sheep with natural scrapie and healthy controls in Norway. J Gen Virol 80, 1073–1077.[Abstract]
van Keulen, L. J., Schreuder, B. E., Vromans, M. E., Langeveld, J. P. & Smits, M. A. (2000). Pathogenesis of natural scrapie in sheep. Arch Virol 16, 57–71.
Woolhouse, M. E. J., Stringer, S. M., Matthews, L., Hunter, N. & Anderson, R. M. (1998). Epidemiology and control of scrapie within a sheep flock. Proc R Soc Lond B Biol Sci 265, 1205–1210.[Medline]
Yu, B., Tiwari, R. C., Cronin, K. A. & Feuer, E. J. (2004). Cure fraction estimation from the mixture cure models for grouped survival data. Stat Med 23, 1733–1747.[CrossRef][Medline]
Received 27 February 2006;
accepted 3 October 2006.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(
(
(
(
(