This paper reports the results of a case-control study of the bovine spongiform encephalopathy (BSE) cases born in Great Britain after the statutory reinforcement of the ban (BARB) on the feeding of mammalian-derived meat and bone meal on 31 July 1996.
A total of 499 suspect clinical cases of BSE, born after 31 July 1996, and reported negative by July 31, 1996 and were compared with the set of 164 confirmed Great BARB cases in Great Britain detected by both passive and active surveillance. Animal-level risk factors (age and type of feed offered) and herd-level risk factors (herd size and type, number of prereinforced feed ban BSE cases born on the holding, the presence of other domestic species and waste management) were obtained for the analysis.
BARB cases were 2.56 times (95 per cent CI 1.29 to 5.07) more likely to be exposed to homemix or a combination of homemix and proprietary feeds were 0.59 times (95 per cent CI 0.50 to 0.69) as less likely to be exposed to the unit increases in the number of prereinforced feed ban BSE cases diagnosed on the natal holding. A supplementary spatial analysis of these cases revealed three areas of excess BARB density: Northwest and Southwest of Wales and Northeast of Scotland.
Statistics from Altmetric.com
BOVINE spongiform encephalopathy (BSE) is a neurodegenerative disease of the cattle first identified in the UK in 1986. A significant outbreak followed and later the disease was detected in other countries around the world. The evidence of the main transmission route of the disease to cattle (Wilesmith and others 1988, 1991) resulted in the introduction of the first control measures in 1988 with the prohibition of feeding of ruminant-derived meat and bone meal (MBM) to ruminants with the exception of milk, milk product or dicalcium phosphate and the declaration of BSE as a notifiable disease (Anon 1988). Although the epidemic was largely controlled with this measure (Wilesmith and others 2000, Stevenson and others 2000a), partial compliance with the specified bovine offal ban, introduced in November 1990 (Anon 1990), and the problems of cross-contamination of cattle feedstuffs with infected MBM (Wilesmith 1996, Stevenson and others 2005) did not totally eliminate the exposure of cattle to the infectious agent. The measures were enforced on July 31, 1996 (the reinforced ban), when the possession and feeding of mammalian-derived meat and bone meal (MMBM) to all farmed livestock, horses and fish became illegal in Great Britain (Anon 1996). In addition, there was a voluntary recall of MMBM and feed-containing MMBM from farms, feed merchants and feed mills. The scheme collected 11,000 tonnes of such material, which was disposed of to landfill at a total cost of £6 million (Hansard 1997). On 1 January 2001, the EU banned the feeding of processed animal protein to farmed animals with certain exceptions (the total feed ban) (Anon 2000).
The first BSE case born in Great Britain after the reinforcement of the ban (BARB), that is, 31 July 1996, was confirmed in 2000. Since the ban until July 31, 2009, a total of 164 cases had been confirmed. Epidemiological reports of the pre-BARB BSE cases (Wilesmith and others 1992, Wilesmith 1996) and of the BARB cases in Great Britain have been published previously (Wilesmith 2002, Wilesmith and others 2005, Burke 2009). An epidemiological update of these 164 BARB cases has been published in a descriptive study complementary to the analyses presented in this study (Wilesmith and others 2010). Despite what is believed to be an extremely high level of compliance with the feed controls since 1996 (Burke 2009), there has been a continued very low incidence of BARB cases for which the aetiology has not been fully elucidated. A feedborne exposure from an exogenous (non-Great Britain) source has been suggested as a potential explanation for the Great Britain BARB cases (Wilesmith 2002). The epidemiological features of the BARB cases identified to date do not provide any evidence to refute this hypothesis (Wilesmith and others 2010).
This study reports the results of a case-control study based on the 164 Great Britain BARB cases detected up to July 31, 2009. The aim of the study was to identify the risk factors for BSE in cattle born after the reinforcement of the ban and to test different aetiological hypotheses, mainly feedborne exposure and environmental exposure through the analysis of potential, putative, animal-level and herd-level risk as well as confounding factors for the presence of BSE.
Materials and methods
Definition of cases and controls
All cases of BSE confirmed in cattle born after July 31, 1996, in Great Britain (hereafter referred to as ‘BARBs’) and detected by both passive and active surveillance by July 31, 2009, were selected as cases. Controls were all suspect clinical cases of BSE, born after July 31, 1996, notified and detected in their natal herd, and for which confirmatory diagnostic methods applied at the time (Gavier-Widén and others 2005) proved negative as at July 31, 2009. Suspicion of BSE in cattle has been statutorily notifiable since June 1988 (Anon 1988). This clinical surveillance was supplemented by active surveillance, which commenced in Great Britain in 2001 and has been amended subsequently to comply with the EU requirements as described elsewhere (Wilesmith and others 2010).
A standard epidemiological questionnaire was completed for all confirmed cases detected by active and passive surveillance and all clinical suspects. This sought relevant case and herd-based information. The feeding history of each case and control animal was summarised for three periods: the first six months of life, from six months of age to the date of first calving and from first calving to death (adult life). The predominant type of supplementary feed consumed during each of these periods was initially categorised as none, homemix, proprietary concentrates, homemix plus proprietary concentrates and unknown. The presence on the farm for the 10 years previous to the birth of the case of pigs, poultry, turkeys, breeding sheep, agisted sheep (typically from November to February) and goats was also recorded. The application of farmyard manure and/or slurry on the farm of interest or from other farms and the application of sewage sludge, blood, blood meal and other abattoir waste products on grassland was each recorded as used, not used or unknown in the 10 years before the birth of the case or the year first present in the herd of exposure. The standard questionnaire also sought information on the use of vaccines and biological products. Their use was so rare that it was not appropriate to include these potential vehicles of infection as putative risk factors in the analysis.
The questionnaire was also completed for the last herd of residence, the natal herd and any intermediary herd, where appropriate, including the date of birth of the animal, the geographical location and the production type of the herd (dairy, beef suckler or mixed). Identifiers of the natal and intermediary herds and the geographical location of the herd of exposure were derived from the British Cattle Movement Service (BCMS) Cattle Tracing System (CTS) database. The most likely herd of exposure for purchased cases was determined by considering the age, duration of stay in the natal and resident holdings, and the average incubation period of the disease.
The year of clinical onset, the year of birth and the number of pre-reinforced feed ban BSE cases on the exposure herds of the study cases and on the natal herds of the study controls were obtained from the main BSE epidemiological database for Great Britain held at the Animal Health and Veterinary Laboratories Agency (AHVLA). The details of all clinical suspects for which BSE could not be ruled out following an official examination together with confirmed cases of BSE detected by active surveillance in Great Britain have been recorded on this database since the identification of the first case of the disease.
Two sets of analyses were carried out. The first provided a description of the spatial distribution of cases, relative to controls. The second quantified the influence of factors associated with an animal's case-control status.
To describe the spatial distribution of BARB cases, two kernel density surfaces representing the number of cattle per square kilometre were constructed using a Gaussian-kernel smoothing function implemented in the sparr package (Davies and others 2011) within R (R Development Core Team 2011). The first (case) surface was based on the location of the herd of exposure of the BARB cases. The second (control) surface was based on the location of the natal herd of controls. These analyses were based on a regular grid of 200 × 200 cells using a fixed bandwidth of 30 km. The logarithm of the ratio of the case surface to the control surface (Kelsall and Diggle 1995) provided an estimate of the spatial variation in BARBs risk throughout Great Britain.
Analyses were conducted using the R package sparr (SPAtial Relative Risk) to test the hypothesis that, after accounting for the spatial distribution of controls, the spatial distribution of BARB cases was uniform across Great Britain. This was done using a procedure based on the calculation of the asymptotic P values assigned to each grid cell of the surface and was based on the z-test (Hazelton and Davies 2009). This approach is an alternative to the computationally intensive calculation of point-wise P values using Monte Carlo simulation (Diggle 2003, Bivand and others 2008). It takes into account the number of observations in an area and gives conservative P values in areas where data are sparse. This analysis allowed us to superimpose asymptotic contours on the log odds surface, delineating areas with an excess of BARB cases, at a specified level of significance, P<0.10 in this case.
To quantify the influence of factors associated with an animal's case-control status, bivariate screening analyses were conducted to test the association between each of the putative risk factors (exposure variables) derived from the questionnaires and BARB status using chi-squared test for categorical variables and t test for numeric ones. A fixed-effect logistic regression model fitted using a backward stepwise approach was applied to select the set of putative risk factors that best explained the probability of being a case. All variables associated with BARB status at an α level of <0.2 at the bivariate level were entered in the model (Dohoo and others 2003). The significance of each explanatory variable was tested using the Wald test. Variables that were not statistically significant were removed from the model one at a time, beginning with the least significant, until the estimated regression coefficients for all retained variables were significant at an α level of <0.05. To account for the correlation in the data at the holding level, the variance-covariance matrix of the model was adjusted using the Huber-White method (White 1982). This method increased the magnitude of the standard error estimates of each of the regression coefficients, making type I errors less likely (rejecting the null hypothesis when it is true). The final model is reported in terms of the estimated coefficients and adjusted OR for each explanatory variable.
Pearson residuals from the logistic regression model were summarised at the holding level (using the technique of Diggle and others 2002) and plotted as a binned omnidirectional variogram using the coordinates defining herd location. This allowed us to quantify the residual, spatially correlated BARB risk at the scales of distance that were small (0 to 50 km) relative to the entire study area.
Model diagnostics were conducted by applying the Hosmer-Lemeshow goodness-of-fit test (Hosmer and Lemeshow 2000), plotting the residuals, visual check of outliers and estimation of the sensitivity, specificity and the area under the receiver operating characteristics (ROC) curve computed using predictions from the model.
To characterise the spatial autocorrelation in residual BSE risk at spatial scales greater than 50 km, the Pearson residuals from the logistic regression model were summarised at the holding level and plotted using the kernel smoothing method described earlier. This surface showed the distribution of holding locations throughout Great Britain with positive-sign and negative-sign residuals. Areas with a predominance of positive-sign residuals were interpreted as areas where the probabiltiy of disease was not explained by the explanatory variables included in the model. Areas with a predominance of negative-sign residuals were interpreted as areas where the probability of disease was less than that predicted by the model.
In order to validate the selection criteria of controls, the subset of all cases sourced by passive surveillance (43) was matched by the surveillance source with 129 controls at a similar ratio of 3:1. A reduced fixed-effects logistic regression model was fitted using these 172 observations. Comparison of the results with the full fixed-effects model is included in the conclusions section.
There were 164 confirmed BARB cases in Great Britain detected by both passive and active surveillance by July 31, 2009. One additional BARB case was confirmed in Great Britain that was born in Republic of Ireland and imported to Great Britain. This case was not included in the analyses presented here. A total of 499 clinical suspect/BSE-negative cases met the selection criterion and were included as controls.
Geographical locations of the holdings of exposure for cases and natal holding for controls are displayed in Fig 1. Fig 2 shows an image plot showing the ratio of the density of BARB cases to the density of controls, expressed on the log scale. This plot showed three areas of relative excess BARB density: the Northwest and Southwest of Wales and the Northeast of Scotland. Distribution of cases and controls by putative risk factor and the results of the bivariate analyses with name and type of the exposure variables, test statistics, degrees of freedom and P values are shown in Table 1.
Table 2 shows the coefficients, adjusted OR and their 95 per cent CI estimated for the fixed-effects logistic regression model. Compared with controls, the odds of feeding with homemix or homemix and proprietary concentrates from birth to six months of age was 2.56 times greater (95 per cent CI 1.29 to 5.07) for BARB cases after adjusting for the effect of the number of pre-reinforced feed ban BSE cases diagnosed on the holding and age at detection. Compared with controls, BARB cases were 0.59 (95 per cent CI 0.50 to 0.69) times as likely to be exposed to unit increases in the number of pre-reinforced feed ban BSE cases diagnosed on the holding of birth. Animals between 4.5 and 6 years were seven times more likely to occur among BARB cases than among controls (95 per cent CI 4 to 12) and being older than six years was 12 times more likely to occur in BARB cases than in controls (95 per cent CI 7 to 21).
The Hosmer-Lemeshow goodness-of-fit test was not significant in the final model showing a reasonable fit (chi-squared test statistic 5.899; 8 df, P=0.659). No outliers in the residuals were influential enough to impact on the coefficients of the covariates in the model when these outliers were excluded from the final model. The area under the ROC curve for the fixed-effects model was 0.82, indicative of a model with reasonable to good ability to discriminate between case and controls on the basis of the parameterised fixed effects.
Kernel smoothed plots of the holding-level residuals model allowed the first-order pattern of ‘unaccounted for BARB risk’ to be visualised (Fig 3). The largest areas of ‘unaccounted for BARB risk’ (areas with aggregations of positive-sign residuals) were the north of Scotland, the west of Wales, the Midlands, an area straddling the counties of Somerset and Dorset, and the southeast of England except part of Kent and East Sussex.
The long incubation period of BSE prevents the assessment of the efficacy of the control measures until at least five years after their enactment, the average time from infection to clinical disease (Anderson and others 1996). In the context of the BARB cases, for which the incidence is very low, suitable analytical epidemiological studies have only been possible after the accumulation of a sufficient number of cases. This prolongs the inevitable delay in the provision of advice on the possible additional control measures to stop further exposure of animals born after the total feed ban implemented across the EU from 2001.
The reduction in the incidence and the risk of infection of BSE in animals born after July 31, 1996 provides further evidence of the beneficial effect of the control measures (Wilesmith and others 2010). Various descriptive epidemiological analyses and a pilot case-control study, based on a more limited number of cases, have been conducted since the occurrence of the first BARB case (Wilesmith 2002, Wilesmith and others 2006). This is the first case-control study of BARB cases conducted in Great Britain, in an attempt to elucidate further their aetiology.
In order to do so, the availability of the BSE-negative clinically suspect cattle, which were born after July 31, 1996, allowed their selection as a group of controls. The details of such suspects and their herds of residence are recorded routinely in the collection of epidemiological data on BSE in Great Britain. The selection of negative-suspect BSE cattle as controls is not ideal given the remote possibility of the misclassification of some of the animals as false negatives by the screening tests, introducing certain ascertainment bias (all controls showed BSE-like signs). The lack of a live-animal diagnostic test, the long incubation period and the low specificity of the clinical profile present a challenge in the selection of controls for any epidemiological field studies of transmissible spongiform encephalopathies (Tongue and others 2009). The approach taken in this study was believed to be the only pragmatic and feasible approach to select a suitable set of controls, presented as such to the Spongiform Encephalopathy Advisory Committee's (SEAC) subgroup in 2005, which was convened to advise on the epidemiology of BARB cases (SEAC 2005).
The selection of controls only sourced from the active surveillance and matched by age (age at detection of controls was significantly lower than of cases) and the time at detection (year) would have overcome the ascertainment bias, but it would have introduced other constraints such as the impossible task of identifying and questioning herd managers on feeding and waste management practices implemented 10 or more years before the case had been confirmed. Even if herd managers could be interviewed, the risk of recall bias and missing data would have significantly compromised the execution of this study design. This is especially so because the herd owner or the herd manager of multiple randomly selected controls would be less motivated to provide the necessary data and information compared with those of clinically suspect controls. This is simply because at the time of the data collection on the clinically suspect controls, there is some likelihood that the animal will be positive on laboratory examination thus stimulating an interest and motivation in the herd owner to provide correct details. Also, a proportion of herd managers of randomly selected controls would undoubtedly be non-respondents.
The final model using only passive surveillance cases and controls included the same variables except the total number of previous cases of BSE. Thus, this is the only variable that appeared to be affected by the potential selection bias of the controls, reassuring the unbiased effect of the feeding system from birth to six months and age at detection. By adding active surveillance cases to the passive ones, the number of previous cases is reduced from 8 to 3.9.
Despite the occurrence of cases and controls during the foot-and-mouth epidemic in Great Britain in 2001 when timely farm visits were not always possible to collect the data, the epidemiological questionnaire was administered for all cases and controls throughout the epidemic. The level of uninformative responses to the question on ‘other species on the holdings’ and ‘waste management practices’ was remarkably low with around 10 per cent ‘unknown’ responses in both cases and controls. In addition, data on feeding practices were only unknown for 5 per cent of the putative exposure herds in the case group. This was as a result of the death of the herd owner or the dissolution of the herds and not being able to trace a person with the necessary information or recollection.
Despite these difficulties in data acquisition, there was no evidence for an association with the presence of other species and the use of extra-farm wastes, the latter being a proxy for environmental contamination, other than the feedborne source, of farms. Information on other potential sources and vehicles of infection were collected routinely. There was no evidence from the detailed field investigations to suspect any of them as risk factors. For example, the rarity of the use of vaccines and biological products resulted in the removal of these potential vehicles of infection from the list of aetiological hypotheses originally considered.
Increases in the number of pre-reinforced feed ban BSE cases reduced BARB risk. It was expected a priori that cases would occur in herds without previous cases with controls more likely to occur in herds that experienced BSE cases earlier in the epidemic. A reason for this could be the potential excess in the notifications of clinical suspects with BSE-like signs by herd managers of positive herds resulting in false positives compared with those with no previous history of BSE, the latter being less prone to notify them. Farmers with a history of BSE could also have taken measures to prevent or minimise the exposure of their cattle by the judicious purchase of feedstuffs, the elimination of the environmental contamination by the thorough cleaning of silos and feed bins, putative reservoirs of the infective agent and/or by the participation in the voluntary recall of MMBM in 1996. Also, unsupervised on-farm burial of clinically affected animals was not common in the early years of the epidemic, when there was only voluntary notification of cases, and therefore this practice represented a limited risk of environmental contamination. Given the above observations, the findings of this study do not provide any evidence of the environmental contamination as a relevant infection route, in line with the available information on the absence of detectable excretion of the BSE agent from infected and affected cattle (Buschmann and Groschup 2005, Everest and others 2006, Espinosa and others 2007).
The use of homemix, or homemix and proprietary concentrates, significantly increased BARB risk. This effect is biologically plausible as the sources of ingredients for homemixing may not be quality-assured as thoroughly as those used for the production of commercial concentrates. Homemixing can result in the use of, and therefore exposure to, larger volumes of individual ingredients in individual diets compared with the composition of proprietary feedstuffs. Previous descriptive analyses have suggested the hypothesis that the origin of BARB cases was the exposure to an exogenous feedborne source (Wilesmith and others 2005, unpublished). Given the strict legislation to control the use of MBM in Great Britain, a likely source of infection of BARB cases is imported contaminated cattle feed ingredients. The results of this study do not contradict the hypothesis of an exogenous contaminated source in the form of feed ingredients used by farmers to produce homemix feedstuffs or by commercial feed compounders.
The significant difference in the age of BARB cases, older than the controls, reflected the natural incubation period of the disease until they become detectable by the surveillance system, and the fact that fewer suspect cases in the older ages were notified in the BARB era as a good proportion of them would go through the active surveillance system, mainly as fallen stock or casualty slaughters.
Studies of the spatial heterogeneity in the risk of BSE for cattle have been notable and informative in understanding the epidemiology of BSE in Great Britain (Wilesmith and others 1991, Wilesmith 1996, Stevenson and others 2000b, Stevenson and others 2005) and in other European countries (Doherr and others 2002, Paul and others 2007). Such analyses have provided the basis for the re-inforcement and changes in the statutory controls to eliminate the feedborne exposure as highlighted elsewhere (Wilesmith and others 2010). In Great Britain, the BSE epidemic can be categorised into three distinct phases based on the geographical change in the risk of infection, in addition to the reduction in incidence in the successive cohorts, following the introduction of the various levels of statutory control.
This study was concerned with the third phase of the epidemic during which the risk was markedly reduced after the re-inforcement of the controls in August 1996 and was, in general terms, more homogeneous geographically than in the previous two phases (Wilesmith and others 2010). The descriptive cohort-based data suggested that in certain areas specific 12-month birth cohorts were at particular risk suggesting a local, but temporary risk of infection. This hypothesis was not open to the analysis because of the small number of cases in each 12-month birth cohort.
The spatial analyses presented in Fig 2 identified three areas of excess BARB case density in the Northwest and Southwest of Wales and the Northeast of Scotland. After controlling for age at detection, the number of prereinforced feed ban BSE cases and the type of feed offered before six months of age (Table 2), the north of Scotland, the west of Wales, the Midlands, an area straddling the counties of Somerset and Dorset, and the southeast of England except part of Kent and East Sussex remained as areas of spatially aggregated, unexplained BARB risk (Fig 3). None of these were areas of unexplained risk in the previous phase of the epidemic (Stevenson and others 2005), providing some support to the hypothesis that exposure was exogenous. That is, neither on-farm contamination from previous cases nor the use of contaminated feed, produced before August 1996, remaining on the farm or in local feed stores appeared to be the major sources of infection. The relatively large size of these areas, together with the relatively widespread distribution of cases throughout Great Britain and in successive cohorts, does not suggest a feedborne source of infection from a single source, assuming that cattle feed is transported over relatively short distances from the place of manufacture to destination herds. Supplementary epidemiological investigations were instigated to investigate the type and sources (ports of entry and source countries) of feed ingredients used for the BARB cases. Although local feed mills, rather than particular feed mills, were associated with cases in some areas (Defra 2005), the length of time between the use of the ingredients and the occurrence of the cases precluded a comprehensive collection of data on the sources of feed ingredients to investigate this aspect further and provide data for this study (J. W. Wilesmith and P. Burke, unpublished information).
In light of the results of this study, the feeding system of animals with homemix, either of national or foreign sources, during the first six months of life, is a risk factor for BARB BSE cases in Great Britain. It is likely that herds with high incidence of BSE were exposed to the infective agent through feedstuff and the environment. However, no environmental contamination via the use of different types of waste on grassland and the presence of other species appeared to be linked to the presence of BSE. This finding together with the observed decreased risk in holdings with the previous cases of BSE reinforce the very low or undetectable risk of infection as result of the persistence of the BSE agent in the environment.
It is not possible to recommend further national controls to prevent new infections. The reinforced control measures implemented across the EU from January 2001 should prevent further exposure of the cattle population, if applied rigorously. Although the long incubation period requires caution in interpreting the effectiveness of the measures, it would, however, be prudent for cattle owners to take care in their selection of feed and feed ingredient suppliers as well as exerting special care in the cleaning and movement of feed bins, silos and other feed storage facilities.
If contaminated feedstuff was responsible for the infection of new cases after the ban, the continuous decline in the incidence of the BARB cases in Great Britain with only two cases confirmed in 2009 should lead to the extinction of the disease unless there are other potential sources of BSE, as yet unidentifiable such as the spontaneous occurrence of BSE or the introduction of a new source of contaminated feedstuff.
This study was conducted under the auspices of the AHVLA project SE0257, funded by the Department of Environment, Food and Rural Affairs (DEFRA). The authors are grateful to Mike Dawson (AHVLA) and Patrick Burke (AHVLA) for their useful comments to this manuscript.
Provenance not commissioned; externally peer reviewed
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.