general enquiries on this form should be made...

General enquiries on this form should be made to:Defra, Science Directorate, Management Support and Finance Team,Telephone No. 020 7238 1612E-mail: [email protected]

SID 5 Research Project Final Report

SID 5 (Rev. 3/06) of 30

NoteIn line with the Freedom of Information Act 2000, Defra aims to place the results of its completed research projects in the public domain wherever possible. The SID 5 (Research Project Final Report) is designed to capture the information on the results and outputs of Defra-funded research in a format that is easily publishable through the Defra website. A SID 5 must be completed for all projects.

This form is in Word format and the boxes may be expanded or reduced, as appropriate.

ACCESS TO INFORMATIONThe information collected on this form will be stored electronically and may be sent to any part of Defra, or to individual researchers or organisations outside Defra for the purposes of reviewing the project. Defra may also disclose the information to any outside organisation acting as an agent authorised by Defra to process final research reports on its behalf. Defra intends to publish this form on its website, unless there are strong reasons not to, which fully comply with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.Defra may be required to release information, including personal data and commercial information, on request under the Environmental Information Regulations or the Freedom of Information Act 2000. However, Defra will not permit any unwarranted breach of confidentiality or act in contravention of its obligations under the Data Protection Act 1998. Defra or its appointed agents may use the name, address or other details on your form to contact you in connection with occasional customer research aimed at improving the processes through which Defra works with its contractors.

Project identification

1. Defra Project code OZ0323

2. Project title

An Integrated Risk-Based Approach to the Control of Salmonella in PIgs in the UK

3. Contractororganisation(s)

Veterinary Laboratories AgencyWoodham LaneNew HawAddlestoneSurreyKT15 1RU

54. Total Defra project costs £ 1,053,974.23(agreed fixed price)

5. Project: start date................ 01 April 2005

end date................. 30 January 2009

SID 5 (Rev. 3/06) of 30

6. It is Defra’s intention to publish this form. Please confirm your agreement to do so...................................................................................YES NO (a) When preparing SID 5s contractors should bear in mind that Defra intends that they be made public. They

should be written in a clear and concise manner and represent a full account of the research project which someone not closely associated with the project can follow.Defra recognises that in a small minority of cases there may be information, such as intellectual property or commercially confidential data, used in or generated by the research project, which should not be disclosed. In these cases, such information should be detailed in a separate annex (not to be published) so that the SID 5 can be placed in the public domain. Where it is impossible to complete the Final Report without including references to any sensitive or confidential data, the information should be included and section (b) completed. NB: only in exceptional circumstances will Defra expect contractors to give a "No" answer.In all cases, reasons for withholding information must be fully in line with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.

(b) If you have answered NO, please explain why the Final report should not be released into public domain

Executive Summary7. The executive summary must not exceed 2 sides in total of A4 and should be understandable to the

intelligent non-scientist. It should cover the main objectives, methods and findings of the research, together with any other significant events and options for new work.Salmonella infections cause up to 50,000 cases of human disease every year in UK. Salmonella Typhimurium (STM) is the second most frequent cause of human disease and this serovar is also found in all domesticated livestock. STM is also commonly isolated from pigs, which may be an important source of some human disease. The aim of this research was to investigate the epidemiology of Salmonella infection in GB pigs and the anticipated efficacy of controls on the prevalence of infection at slaughter and the number of human cases of salmonellosis. Each main research objective is summarised below.

Objective 1 – Epidemiological studiesA literature review showed that control of Salmonella infection in pigs could be achieved by reducing risk of introduction of infection; reducing transmission; reduced mixing of pigs and isolation of sick pigs; stimulating an acid gut environment; or by improving pigs’ resistance to infection e.g. by vaccination or control of other disease e.g. PMWS or PRRS. However, none of the measures are applicable in all circumstances. Introduction of Salmonella infection with weaners is an important risk factor but there are no accredited sources of Salmonella-free weaners in GB.

A literature review was conducted to investigate the proportion of human salmonellosis cases that might be attributed to pigs or pig products – the Population Attributable Fraction (PAF). Only 23 of 700 articles enabled estimation of PAF. Two considered sporadic human cases and did not show any association with pigs or pig products. Four studies considered environmental exposure and three showed an association between animal contact and human salmonellosis but the specific contribution of pigs was not measured. The remaining studies were of outbreaks and these did show a clear association between human disease and pig products, e.g. ham or salami. Most human cases are sporadic.

Data from long term studies on 21 farms showed contact with faeces increased risk of Salmonella infection and that older sows were less likely to be infected than younger pigs. During the study, more than 9,000 Salmonella isolates were found and more than 50 different serovars were identified. The most frequent were STM and S. Derby, which is the second most frequent serovar found in GB pigs. Salmonella was also isolated from insects, arthropods, rats, mice and other wildlife as well as other domestic species. The effectiveness of an oral STM vaccine administered via drinking water to 20 batches of finishing pigs was evaluated. No formal control group was tested but results were compared to GB survey data. Although fewer pen samples were positive than encountered during a survey of 107 farms (in project OZ0316), the MJ ELISA results, prevalence of infection in lymph nodes and in caecal contents at slaughter showed no significant difference from unvaccinated pigs in the general population. MJ ELISA results from ZAP/ ZNCP were related to farm-level information recorded by QA schemes.

SID 5 (Rev. 3/06) of 30

Additional information was collected via questionnaires. Initial analysis of these data showed that high prevalence MJ ELISA-positive farms in Yorkshire and Humberside were clustered together and that pigs housed on slatted floors were less likely to be MJ ELISA positive.

Objective 2 – MicrobiologyOn 4 farms pens of pigs were monitored from approximately 10 weeks of age through to slaughter. Salmonella isolates were tested by Pulsed-Field Gel Electrophoresis (PFGE) and Variable Number Tandem Repeat (VNTR) analysis, which enabled detailed discrimination between strains. Results showed that specific VNTR types persisted on some farms and could also be detected at slaughter. Other Salmonella strains that had not been found on farms but had been isolated from lairage pens at the abattoir were also found after slaughter. VNTR and PGFE were applied to more than 700 STM isolates from pigs, human cases and other domesticated species to investigate whether these techniques might characterise the origin of strains from human illness. Cluster analysis revealed 4 large closely-related groups, some of which were only found in pigs and humans. These results corroborate the role of pigs as a source of some human STM cases but do not yet allow accurate attribution.

Samples of caecal content, lymph node and carcass swabs were tested to estimate the number of Salmonella bacteria which they carried. Carcass swabs had a much lower weight of infection than caecal content or lymph nodes. These data were used to inform the abattoir module of the QMRA.

Objective 3 – Quantitative Microbial Risk Assessment (QMRA)A mathematical model of transmission of Salmonella amongst breeding pigs and their offspring was added to an existing QMRA. The model predicted that a Salmonella- free farm had a 60% probability of becoming infected if a single infected pig were introduced. It also predicted that one third of pigs in all age groups on a farm could be excreting Salmonella at one time and that a quarter of pigs would be Salmonella carriers when slaughtered. Objective 4 – modelling STM infectionSample data from 48 farms were analysed to estimate between and within pen transmission rates. With respect to salmonella infection, any particular pig may be naïve and fully susceptible, infected or have recovered from infection. The transmission rates estimate how rapidly pigs change from one state to another and knowledge of these is important in designing control measures. The results were also used to estimate the number of pooled faecal samples that would be required to detect Salmonella at different prevalences in pig herds of varying sizes.

Objective 5 – spatio-temporal analysis of ZAP dataData from ZAP were used to investigate the effects of season and location upon the risk of Salmonella infection. Local “hotspots” of infection were found and some seasonal trends were observed in areas with a greater density of pig holdings (e.g. Yorkshire). These holdings were more likely to change their status from negative to positive than holdings in SW England, where the density of pig holdings is lower.

Objective 6 – socio-economicsCosts were calculated for interventions including rodent control, feeding with organic acids and vaccination at the farm level and vertical scalding, bunging and carcass washing at the abattoir. A cost-effectiveness ratio was estimated for each intervention by combining the cost with the predicted effectiveness estimated by the QMRA.

Data to estimate the costs associated with human STM disease were collected through questionnaires sent by the Health Protection Agency (HPA) between July and December 2008. Direct expenses included costs of transport, medicines and loss of days of work. The public health costs were estimated separately for the families of the cases and the NHS. The most frequent symptoms were diarrhoea and abdominal pain. 94% of STM patients were sick on average for 13.5 days. 69% of STM cases needed someone to care for them whilst they were ill; 23% of carers took a day off work.

The mean direct family cost for STM was £54.91; the indirect cost due to loss of income was £182.35. Estimated average NHS costs per patient were £611.44. These results were linked to the QMRA. We assumed that only 1 in 4 of the predicted cases are reported and the cost of an unreported case is only the average family direct cost. The results of the integration of the MRA and the cost of human illness were used to predict that the annual cost of Salmonella attributable to pig meat in UK is £234,276. The QMRA predicts that interventions may reduce the incidence of human STM illness by 12% -70% but the annual costs of these interventions exceed the value of the reduced human caseload. Therefore, it is more relevant to consider the cost-effectiveness of control strategies rather than the cost-benefit ratio. Overall, abattoir interventions and especially bunging were predicted to be the most cost-effective.

SID 5 (Rev. 3/06) of 30

However, these results depend on theoretical costs and predictions of efficacy made by mathematical models alone. The key message is that the QMRA can be used to make these predictions, which will be extremely important to decision-makers. The validity of these predictions depends upon obtaining real field data to provide values for the model parameters. Finally, since the QMRA is a farm-fork model, it may also be used by different stakeholders to arbitrate costs. The model also offers the potential to differentiate farms according to risk factors, so that the cost-effectiveness of different interventions according e.g. to farm management might be considered.

Objective 7 – integrationSocio-economic data were incorporated into the QMRA to estimate the predicted number of cases and annual costs from human salmonellosis attributable to pigs. We predicted that 500 cases might arise each year at a cost of about £250,000. We also used the QMRA to consider whether accounting for different areas of the UK that presented differing prevalence of MJ ELISA positive pigs might influence the predictions from the model. This was not the case; however, inclusion of spatial effects does offer the opportunity to test whether different interventions might be more cost-effective in different areas or to test the possible impact of focusing interventions within high-risk areas.

Objective 8 – decision support tool (DST)A prototype user interface for a DST was developed during the project. However, the change from ZAP to ZNCP in 2008 rendered the particular tool obsolete and it was agreed that further development be suspended. The experience gained would be useful if such a tool were required in the future – e.g. once the nature of any control plan required by the EU is known.

Objective 9 – communicationIn the longer run, Salmonella control must be included in herd health and not seen as an optional extra. If this is to be achieved, it is essential that private veterinary consultants to pig producers become engaged in the process. Scientific outputs from our work have already been presented to industry and to vets. However, science alone is not sufficient and it is important to work at understanding the perceptions of stakeholders especially in farms and abattoirs if farmers are to be motivated to change behaviour to bring about the adoption of sustainable control plans.

Conclusions1. Salmonella infection in UK pigs causes some human illness but this has not been quantified2. Control measures may be effective on some farms or abattoirs but no universal solutions have

been found. 3. The average cost of a reported human case is around £850. We predict that 500 cases may be

attributable to infection from pigs each year, at an annual cost of approximately £250,000.4. The most cost-effective interventions are predicted to be in abattoirs. However, the cost of any

effective intervention exceeds the predicted economic benefits.5. Current model outputs should not be assumed to be accurate but the QMRA can be used to

model different scenario’s that might be defined by policy-makers in Government or industry.6. It is essential that control measures are commensurate with risk, to avoid demanding that industry

accepts an economic burden yet society gains no net benefit from their investment.

Future work1. The most critical data that are not currently available to decision-makers concerns the proportion

of sporadic human salmonellosis cases that may be attributed to pigs (PAF). This could be estimated, for example, through a case-control study.

2. For many farms, the greatest risk of introducing Salmonella infection lies with the introduction of live pigs, especially weaners. Research should be directed towards the “Salmonella-free weaner”.

3. Further studies on the costs and efficacy of on-farm and abattoir interventions are necessary.4. Further studies are needed on the potential for vaccination to control Salmonella in pigs.5. There will be an on-going need for surveillance for Salmonella in pigs and mathematical models

could be utilised to optimise approaches, for example with respect to sample type, sample size and diagnostic test.

6. Some husbandry systems that are perceived to be welfare friendly, e.g. outdoor rearing are also associated with an increased risk of infection. Stratification of our QMRA by system would enable an estimation of the overall contribution of these systems to the public health threat to be analysed.

Project Report to Defra

SID 5 (Rev. 3/06) of 30

8. As a guide this report should be no longer than 20 sides of A4. This report is to provide Defra with details of the outputs of the research project for internal purposes; to meet the terms of the contract; and to allow Defra to publish details of the outputs to meet Environmental Information Regulation or Freedom of Information obligations. This short report to Defra does not preclude contractors from also seeking to publish a full, formal scientific report/paper in an appropriate scientific or other journal/publication. Indeed, Defra actively encourages such publications as part of the contract terms. The report to Defra should include: the scientific objectives as set out in the contract; the extent to which the objectives set out in the contract have been met; details of methods used and the results obtained, including statistical analysis (if appropriate); a discussion of the results and their reliability; the main implications of the findings; possible future work; and any action resulting from the research (e.g. IP, Knowledge Transfer).

Project objectives

1. Epidemiology of Salmonella Typhimurium in GB weaner production1.1 Review evidence for the attribution of human disease with Salmonella Typhimurium in GB to pig

exposures (including consumption of pig meat, cross-contamination, occupational exposures and environmental exposures).

1.2 Epidemiological analysis of data from a longitudinal study of 21 pig farms between 1995-2004 to investigate:

Prevalence of infection and temporal variation in prevalence by age group Persistence of STM strains within farms Associations between infection in weaners and finishers

1.3 In collaboration with stakeholders, propose and test potential interventions to control STM infection in weaners and determine the impact upon STM infection and prevalence of MJ ELISA positive pigs

1.4 Develop a questionnaire for herds monitored by the ZAP programme to supplement quality assurance scheme data with information on risk factors for |STM infection

1.5 Combine MJ ELISA data, QA data and questionnaire data for epidemiological analysis1.6 Evaluate the impact of scheduled slaughter, determined by ZAP score, upon the prevalence and

burden of STM carcass contamination2 Microbiology

2.1 Apply appropriate strain typing methods to selected STM isolates of known provenance in order to:

Examine persistence of strains on farms and through abattoirs Compare human and pig isolates Investigate transmission between and within pens

2.2 Use semi-quantitative methods to estimate infectious burden in the farm and abattoir environment, in individual pigs and on carcasses

3 Farm to fork microbial risk assessment (MRA)3.1 Review scientific and other literature published since July 2003 and update MRA accordingly3.2 Develop a stochastic breeder-weaner model for STM infection and incorporate it into the MRA3.3 Evaluate other methodological approaches for components of the MRA especially where data are

sparse, e.g. Bayesian inference and modify MRA as appropriate3.4 Incorporate results from current and planned research and surveillance activities

4 Modelling STM infection4.1 Estimate parameters for between and within-pen spread for the farm module of the MRA4.2 Develop an approach to estimate herd-level prevalence from pooled samples

5 Spatial and temporal analysis of ZAP data5.1 Assess spatial and temporal variation in the ZAP database and examine covariate relationships5.2 Modelling the probability of a premises changing infection status between visits5.3 Use of MJ ELISA data to predict areas with an increasing risk of Salmonella infection

6 Socio-economics6.1 Develop least-cost models for interventions at farm and abattoir level within the MRA6.2 Investigate economic optimisation of interventions along the food chain to reduce human STM

illness6.3 Estimate the public health costs of salmonellosis due to STM6.4 Conduct a full farm to fork stakeholder analysis

7 Integration7.1 Extend MRA to include spatial elements derived from 5.1 – 5.37.2 Use results from objectives 1, 4 and 5 to review and revise parameter estimation in MRA

SID 5 (Rev. 3/06) of 30

7.3 Use results from socio-economic studies (6.1-6.4) into MRA to allow concurrent modelling of the physical impact of possible interventions and the economic costs and benefits

7.4 Use the model in scenario analysis8 Decision support tool (DST)

8.1 Review decision support tools and relevant literature to inform design and delivery8.2 Reduce the run time for the farm module of the MRA8.3 Create a prototype CD to provide a stand-alone tool when loaded into a personal computer8.4 Evaluate the DST with stakeholders

9 Communication9.1 Evaluate opportunities to enhance Salmonella control through existing programmes9.2 Promotion of Salmonella control.

Some of these objectives were subject to revision during the project, due in part to identification of cross-cutting issues with other Defra and FSA projects. The following sections summarise project activities under each heading.

BackgroundThere are about 14,000 reported human cases of salmonellosis every year in Great Britain (GB) and probably three times more in the community. Most are due to Salmonella Enteritidis, which is associated with poultry. The second cause of human illness is Salmonella Typhimurium, which is isolated from all domesticated livestock. Abattoir surveys of GB pigs showed that about ¼ carried salmonella, particularly S. Typhimurium (STM), in their gut at slaughter. The British Pig Executive (BPEx), supported by the Food Standards Agency, launched the Zoonoses Action Plan for Salmonella in Pigs (ZAP). ZAP utilised a meat juice enzyme-linked immunosorbent assay (MJ ELISA) to detect antibodies against salmonella. Presence of antibodies indicated that the pig had been infected with salmonella at some time but did not necessarily imply infection at slaughter. The results were used to designate a ZAP score for each holding and if more than 50% of pigs were MJ ELISA positive then action was needed. Individual farms reduced their ZAP score but there was no change in the prevalence of MJ ELISA positive pigs in GB and in 2008, ZAP was replaced by the Zoonoses National Control Plan (ZNCP). The European Commission (EC) Zoonoses Directive places responsibility for food safety upon primary producers. A baseline survey for Salmonella in slaughter pigs was conducted in 2007. This confirmed that a quarter of UK pigs were infected at slaughter and showed that UK was one of the worst countries in the EU. The EC will set targets for control of salmonella in pigs after a quantitative microbial risk assessment (QMRA) and cost-benefit analysis are completed.

Control of salmonella in pigs is challenging. Salmonella grow very successfully in the gut. They also survive for long periods in the environment. Presence of other infections, such as post-weaning multi-systemic wasting syndrome (PMWS) or porcine respiratory & reproductive syndrome (PRRS) are also associated with an increased risk of infection. Some management systems, such as outdoor production or solid floors with straw, may also increase the risk of salmonella infection. Policy-makers and the pig industry must have strong evidence base to ensure that the anticipated impact of proposed control measures is understood and that the costs of these measures are commensurate with the benefits from a reduced public health burden. The aim of this research was to investigate the epidemiology of salmonella infection in GB pigs and the anticipated efficacy of controls on the prevalence of infection at slaughter and the number of human cases of salmonellosis. Each main research objective is summarised below and these are related to primary project milestones..

OBJECTIVE 1 Epidemiology of Salmonella Typhimurium in GB weaner pig production

1.1 Review evidence for the attribution of human disease with Salmonella Typhimurium in GB to pig exposures (including consumption of pig meat, cross-contamination, occupational exposures and environmental exposures).A systematic review of studies that reported the association between exposure to pigs or pig products and human illness due to Salmonella Typhimurium (STM) was conducted. Although certain strains of STM, for example U288 and U302, are specifically associated with exposure to pigs, there is no quantitative estimate of the risk and the proportion of all human STM cases attributable to exposure to pigs. The relative importance of environmental exposure or consumption of contaminated food is not known either. Many strains of STM may be found in foodstuffs including beef, dairy products, lamb, eggs, chicken and turkey as well as a range of other produce, e.g. peanut butter, which have been cross-contaminated. A systematic search was conducted of published and unpublished sources to identify all literature that reported a quantitative estimate of the association between STM in humans and exposure to pigs. We considered papers which did not offer a quantitative measure (e.g. time changes in STM strain types in humans and in pigs in routine data), but used them only to inform the discussion. The PubMed and Embase databases were searched. ISI Proceedings, SIGLE and a number of potentially relevant (e.g. HPA, VLA) websites were examined for unpublished sources of information. Articles were grouped in order to examine comparable results within each of the following categories:

SID 5 (Rev. 3/06) of 30

1. whether the study was of pig consumption, occupational exposure, or environmental and recreational exposure;

2. whether the cases included in the study were identified as belonging to an outbreak or not (‘sporadic’) Outbreaks were further categorised into:

a. those which could be linked to a point source, e.g. a catered function, b. those which were identified through a sharp increase in the number of laboratory confirmed

cases of a particular phage type during a short period of time 3. whether the case definition of STM included laboratory confirmation; 4. whether the study was conducted in Great Britain or another developed country; 5. the specific phage type of STM studied.

The odds ratio (OR) or risk ratio (RR) was generally presented in papers which estimated the magnitude of the association between exposure to pigs and risk of STM directly, and in instances where it was not presented, it could be calculated from the distribution of cases and controls. Similarly, the population attributable fraction (PAF) could be calculated in instances where the distribution of cases and controls and the prevalence of exposure in either the cases or controls were presented.

The interpretation of the PAF is very different for point source outbreaks, laboratory-detected outbreaks and for sporadic cases. For outbreaks in which the timing and place of exposure are known e.g. occurring among people who attended a social event, the PAF is the proportion of the cases in the outbreak which are attributable to the consumption of a particular food. For analyses of outbreaks which are identified because of a sharp time limited increase in the number of laboratory confirmed cases of a specific serotype, the PAF corresponds to the proportion of cases during the period of the outbreak which are attributable to consumption of that food. In studies of sporadic cases the PAF represents the proportion of all such cases that are attributable to a particular product. PAF is a much more useful measure than the RR because the RR in studies of sporadic cases will vary depending on whether we investigate the RR associated with a general food (“all sausages”) or the culprit (sausage brand x). The PAF, however, is the same whichever approach we take because the PAF depends on the proportion of the population that consumed the food and on the RR associated with that food.

A total of 787 articles were identified but examination of the abstracts revealed that 676 of these were irrelevant to the research question, leaving 111 of interest. A further 7 articles were found through examination of the references listed in the selected articles or from personal knowledge of the researchers. Of the 118 articles, 23 papers provided some direct quantitative estimate of the association between exposure to pigs and risk of STM. The remaining 95 papers were relevant to the research question but did not quantify the association directly. Almost all of the studies were of STM cases that had been identified as part of an outbreak, either because of a sharp increase in the number of laboratory confirmed cases or because the time and place of exposure was known. Only 3 studies included cases of sporadic disease. There were 15 case-control studies and 4 studies used cohort methods in which the subsequent risk of STM among persons attending a catered function was compared between those who ate pork and those who did not. Three studies used mathematical modelling techniques and one study was microbiological. Seven studies examined all STM cases regardless of phage type and two studies included serotypes other than STM. The remaining 14 studies were analyses of particular phage types. Eleven of the 23 studies were conducted in Great Britain. No articles presented a PAF for the association between exposure to pigs and STM risk but these could be calculated from the presented data for most studies.

Only two studies examining consumption of pig products were of sporadic cases of disease, neither of which provided evidence of increased risk associated with pork consumption. In contrast, 15 studies analysed outbreak cases of STM and provided strong evidence that consumption of pig products was often the cause. Seven of these were instigated because of an increase in the number of laboratory reported cases. Two reported increased risk associated with ham consumption, for phage types DT124 and DT194 respectively. For both outbreaks, the proportion of cases attributable to the consumption of ham was in the region of 60-70%. A third study found strong evidence of an association (p<0.001) between the consumption of salami sticks during a 7-day period and risk of DT124. There were 8 studies in which the time and place of a catered function were identified and these also produced convincing evidence that consumption of pork or pig products was likely to be the cause of the increase in STM cases. The odds ratios of these studies generally varied from about 2-fold up to 25-fold and the PAF’s ranged from 40-90%. However, it should be noted that if there had been a high attack rate in any outbreak, then the Odds Ratio would be more extreme than the Risk Ratio and this will over-estimate PAF.

Four studies investigated the association of STM in humans with environmental and recreational exposures. Three of these found evidence of associations, although not specific to pig exposure. Findings included a 5-fold increase in risk of STM associated with contact with ill animals and farm animals and increased risks of both sporadic DT-104 and sporadic non-DT104 cases associated with living on a farm. However, despite the relatively large risk ratios in some studies, the estimated PAFs were low (10%-20%), since the probability of exposure was low. One study examined the association between occupational exposure and risk of STM but found no evidence of a difference in antibody levels between slaughterhouse and greenhouse workers.

SID 5 (Rev. 3/06) of 30

Overall, there was clear evidence of the effect of consumption of pig products in causing some outbreaks of STM. The evidence for the role of pig products in causing sporadic cases is less compelling. Some environmental and recreational activities are associated with increased risk of sporadic cases of STM in humans; but it is not clear how much of this is due to environmental exposures to pigs or to other animals. No summary effect for the association between STM and pig exposure was estimated firstly, because one estimate is required for outbreaks and another for sporadic cases. The most interesting measure is the contribution of pigs to sporadic cases in humans, as these form the greatest majority of human cases, but there was only one study in England and this only explained a small proportion of cases, with more due to environmental exposures than to consumption of meat. Secondly, although there are a substantial number of studies of outbreaks, not all outbreaks are investigated. Furthermore, the search criteria were intended to identify papers that did associate STM with pigs or pig products and we did not seek studies that might have identified associations with other foods. Finally some studies did not take account of confounders and used univariate analyses while some used multivariate analyses. An accurate estimation of PAF is an essential pre-requisite for socio-economic studies of the benefits for controlling Salmonella infection in pigs, since the societal benefit is a reduction in human cases. Further data are urgently needed, as cost-benefit analysis will inform EU targets for control.A draft report was circulated in September 2006, satisfying project milestone 01/01. A scientific paper is under preparation for submission to a peer-reviewed journal

1.2 Epidemiological analysis of data from a longitudinal study of 21 pig farms between 1995-2004 to investigate:

Prevalence of infection and temporal variation in prevalence by age group Persistence of STM strains within farms Associations between infection in weaners and finishers

Prevalence of infection and temporal variation in prevalence by age group.A total of 34,805 samples were collected from 21 pig farms that were monitored for varying periods between May 1995 and March 2004. These farms were originally identified in a previous Defra project (OZ0134) and monitoring was continued under OZ0316. The majority of samples were from pigs but a large number of environmental samples were also taken. Most pig samples were pooled, although rectal swabs from individual pigs were also collected.

Salmonella was isolated from 27% of all samples collected and all farms were positive at some stage. The proportion of samples that yielded Salmonella from areas contaminated with faeces was 57.9% compared to 29.4% of pooled faeces and 8.1% of feed samples. It should be noted that these samples were not selected at random and this observation should not be interpreted as suggesting that 8% of feed contains Salmonella. At the univariable level, using only pooled faecal samples there was a significant difference in the prevalence of Salmonella and STM between the different pig types, with gilts and growers having the highest prevalence at around 37% each and sows having the lowest at 22.5% (Table 1).

Table 1. Prevalence of Salmonella and STM in pooled faecal samples by pig type Pig type Number of pooled faecal

samplesPrevalence of Salmonella Prevalence of STM

Weaners 2697 35.1% 19.5%Growers 3805 37.5% 26.2%Finishers 3588 30.1% 18.6%Gilts 1568 37.9% 18.0%Sows 6397 22.5% 3.8%Boars 1062 31.5% 11.4%

When controlling for farm type there was a significantly lower prevalence of Salmonella in sows than in weaners, growers or gilts although no other differences among the pig types were detected including no significant difference between weaners and growers or finishers. The dataset did not accurately record the age of the pigs sampled beyond broad classification of weaner, grower etc. Due to the inconsistency of sampling on these farms temporal analysis could not be carried out but some temporal analysis is discussed later. These limitations demonstrate that intensive sampling which is not based on random sampling has limited epidemiological value since it cannot be formally analysed nor can findings be extrapolated to a wider population. Nonetheless, these case studies are useful for generating hypotheses and testing sampling methods. They also provide valuable insights into infection on individual farms.

Persistence of STM infection within farms.A total of 9392 isolates from the 21 farms were serotyped and 50 different serovars or partial types were identified. Salmonella Typhimurium was isolated most frequently (51.6% isolates) followed by S. Derby (17.0%),

SID 5 (Rev. 3/06) of 30

which reflects findings from other surveys and scanning surveillance. The most common phage types of S. Typhimurium that were encountered were DT104 and U302.

In addition to the pigs, Salmonella was also isolated from insects (flies and beetles) mice, wildlife and other domestic species on these farms. In particular, Salmonella Typhimurium was found in one or more samples from arthropods, flies, badgers, foxes, mice, rats, wild birds, cattle, cats, dogs, poultry and sheep.

Various interventions were employed on 8 of the study farms. To maximise the available data, and power to the analysis, the persistence of all Salmonella serovars was analysed, not just STM. Five of these farms did show a reducing Salmonella prevalence of infection with time whilst 2 farms showed an increased prevalence after intervention and 1 farm was unchanged. There were no control groups on these farms and the farms that did not intervene are an inadequate control group, since they were monitored for different periods of time. Thus, it is impossible to formally test whether these interventions were effective. However, the experience gained in this study proved useful in planning a later intervention study (OZ0316).

Associations between infection in weaners and finishers.There was considerable variation in the prevalence of infection within individual farms and between farms during the study. In this study, the prevalence of infected pooled faecal samples was significantly greater in growers and finishers from breeder-finisher farms (9 farms) than those sampled on specialist finisher farms (6 farms). However, since these farms are not representative of any larger population, this finding cannot be assumed to be generally true. As discussed above, when controlling for farm type there was no significant difference between weaners and growers or finishers.

Conclusions: Salmonella was widespread and persistent on these farms during the course of this study, although there was considerable variation in prevalence within and between farms over time. The study was not designed to allow formal statistical testing of the impact of interventions but the results do illustrate the necessity of control groups for this purpose, since no clear general conclusions could be reached. Salmonella could also be found amongst a diverse range of species on these farms, illustrating the potential for these species to act as reservoirs of infection for the particular farm and as potential agents of dissemination to other farms or areas. The non-representative nature of the farm selection and the sample selection within farms precluded more detailed epidemiological analysis and limits the extent to which these results can be extrapolated to any wider population. However, it must be emphasised that this was not the intent of the original study and the knowledge gained through in depth study of a few farms complements results from later, more formal approaches in a follow-up project (OZ0316).A draft report on these analyses was circulated in December 2006, satisfying project milestone 01/02.

1.3 In collaboration with stakeholders, propose and test potential interventions to control STM infection in weaners and determine the impact upon STM infection and prevalence of MJ ELISA positive pigs.Commercial pig producers face ongoing problems in controlling Salmonella in finishing pigs. Veterinary consultants representing two integrated pig production enterprises independently approached VLA for advice on assessing the efficacy of Salmonella vaccines to reduce the prevalence infection in pre-slaughter pigs following reported success with the use of a Salmonella Typhimurium oral vaccine that is licensed for poultry. It was the judgement of these practitioners that vaccination was justified as a possible method to reduce the risk of clinical disease in the pigs, minimise any potential risk to public health and resolve quality assurance issues consequent upon a high “ZAP” score. The results of the sampling carried out to determine the effect of vaccination on Salmonella levels in pigs prior to slaughter is outlined below. Recruitment of these farms satisfied project milestone 01/04.

MethodsThe vets each selected 10 farms for administration of the vaccine; VLA was not involved in the selection of the farms, only in the monitoring of the results. On each farm, an oral Salmonella vaccine was delivered via the drinking water to all pigs on two occasion’s approx 3-5 weeks apart. No control farms were selected and there were no control batches of pigs on the vaccinated farms. The study was carried out from to March to October 2007.

SamplingOn each farm, pooled pen floor faeces from 30 pens were collected by the veterinarians after restocking and prior to administration of the vaccine. At slaughter, 50 MJ ELISA samples, 50 caecal content samples and 12 lymph node samples were collected from batches of pigs from each of the farms to test for antibodies against Salmonella and for microbiological culture to isolate Salmonella bacteria. Abattoir visits were carried out by VLA staff.

Analysis

SID 5 (Rev. 3/06) of 30

Faecal samples: As no control was available a previous study is used for comparison. Pooled faecal samples taken during project OZ0316, a cross sectional study on 107 randomly selected finishing pig farms, was used to provide an estimate of average population prevalence. Abattoir samples: To obtain comparable estimates of the levels of salmonella in the population, the results of the EU Salmonella in fattening pigs survey was used (VLA, 2008). The survey sampled 641 pigs in 18 UK abattoirs between October 2006 and September 2007.

A logistic regression model with clustering of samples by farm was used. The outcome was at the pig level, positive or negative for salmonella and the population/study was included as a covariate. This enabled clustering of pigs from the study farms to be taken into account. Unfortunately the pigs sampled during the EU survey could not be assigned to farms and these were treated as a single cluster.

Results

1. Faecal samples: The average proportion of positive pens on each farm prior to vaccination is shown in Table 2. In total 7.3% of the pens were positive but this ranged from 0 to 30% of positive pens on the farms. In contrast, the mean Salmonella pen prevalence in OZ0316 was 22.3% (range 0-97%). This was a statistically significant difference, with samples from vaccinated farms less likely to be positive for Salmonella (OR 0.30, p=0.034). 2. Salmonella at slaughter: At slaughter, all batches of pigs were sampled. meat Juice, lymph nodes and caecal contents were collected and tested. 3. Meat Juice: The prevalence of salmonella in meat juice samples in the UK, as determined during the EU survey for Salmonella in fattening pigs, was 25.4% (163/641). The overall prevalence of salmonella in the vaccinated study pigs was 36.2% (range 7.3 -86.3%). There was no significant difference between the two populations (OR 1.7, p=0.149).4. Lymph nodes: Salmonella was isolated from 21.8% (139/639) lymph node samples in the EU survey, and this was not significantly different to the prevalence 24.1% in the vaccinated study pigs (OR=1.42, p=0.33).5. Caecal contents: Overall, 27.9% (range 4.0-92.3%) of the pigs from the vaccinated study farms had Salmonella isolated from the caecal contents. This is not significantly different to the results of the EU survey in which 21.9% (137/626) pigs were positive (OR=1.4, p=0.518). The results are shown in Table 2.

Table 2. Results from sampling of vaccinated study farmsPooled pen faecal samples Meat Juice Lymph nodes Caecum contents

Farm

Pig

age/

wt

Num

ber

sam

pled

No.

pos

itive

% p

ositi

ve

Sero

type

Num

ber

sam

pled

No.

pos

itive

% p

ositi

ve

Num

ber

sam

pled

No.

pos

itive

% p

ositi

ve

Sero

type

Num

ber

sam

pled

No.

pos

itive

% p

ositi

ve

Sero

type

1 40kg 30 0 0.0 51 1835.

3 12 1 8.3 1 STM 52 4892.

3

1 Bovis morbif, 5 Derby, 9 London, 33 STM

2 8kg 24 2 8.3

1 Reading, 1 STM 51 4 7.8 52 12

23.1 12 STM

3 55 4 7.3 12 0 0.0 50 714.

05 Derby, 2 STM

4 25-30kg 30 1 3.3 1 STM 51 2141.

2 12 866.

7

7 London, 1 STM 50 16

32.0

3 London, 13 STM

5 8 wks 30 413.

3 4 STM 50 1020.

0 12 1 8.31 Reading 51 11

21.6

2 Bovis morbif, 9 STM

6 28kg 30 930.

0 9 STM 49 1326.

5 12 325.

01 Derby, 2 STM 50 2 4.0 2 STM

79wks/22kg 30 0 0.0 51 25

49.0 12 2

16.7 2 STM 50 2 4.0 2 STM

8 46 1021.

7 12 325.

02 Derby, 1 STM 50 2 4.0

1 Derby,1 Reading

9 51 4486.

3 12 541.

7 5 STM 50 3876.

01 Reading, 37 STM

10 40kg 30 0 0.0 50 3468.

0 12 325.

0 3 Derby 51 3 5.9 3 DerbyTotal

204

16 7.8

505

183

36.2

108

26

24.1

506

141

27.9

SerotypesThe most frequent serotype to be isolated from the positive samples was S. Typhimurium at 75.4% of samples. This was higher than was found in the EU fattening pig survey where 50.7% of positive samples were S. Typhimurium. This strongly supports the findings presented above and suggests the vaccine had no additional effect on the prevalence of S. Typhimurium.

SID 5 (Rev. 3/06) of 30

ConclusionsThis study was set up in response to a request from a pig company to test the efficacy of a vaccine for Salmonella in finishing pigs. The study was reliant on the co-operation of the industry and good will of the pig producers taking part and although VLA was able to advise on the samples that were taken, there was no scope for including control batches of unvaccinated pigs or unvaccinated farms in the study. This means that the conclusions that can be drawn regarding vaccine efficacy are limited. All 10 farms practised all-in/ all-out management; the pig accomodation was completely emptied, cleaned and disinfected between batches.

Comparison of the pooled floor faecal samples from pens on the farm and from those taken from a random selection of UK farms suggest the study farms may actually have started with a lower than average Salmonella prevalence. Despite this, at the time of slaughter there was no significant difference between the pigs that received the Salmonella vaccine and pigs sampled as part of the EU fattening pig survey.

As shown in Table 2 even at slaughter a high proportion of the positive samples were S. Typhimurium, and it was decided that statistical analysis of this serotype alone was not waranted. In general, the serotypes present in the study farms were comparable to those found in the EU survey with a high proportion of STM and S. Derby.

In conclusion, the results presented here suggest that the vaccine that was used on these study farms had no effect on the prevalence of salmonella in pigs at slaughter and the vaccinated pigs had prevalence levels comparable to the general pig population. However, it should be remembered that no control farms or batches of pigs were used and therefore these results must be interpreted with caution.

A second vaccination study was initiated at the request of a breeder-finisher and his veterinary consultant. Alternate groups of sows were either vaccinated with an attenuated STM strain (Salmoporc ®) and their progeny are being followed to slaughter. This study, which is also supported by BPEx, will conclude later in 2009.

ReferencesVLA (2008) FINAL REPORT “Commission Decision 2006/668/EC concerning a baseline study on the prevalence of Salmonella in slaughter pigs sampled in abattoirs” and further work on Salmonella and Campylobacter from sampled pigs. 1st October 2006 to 30 September 2007 UNITED KINGDOM. VLA Project Code FZ2023 FSA Project Code VFDS001.

1.4 Develop a questionnaire for herds monitored by the ZAP programme to supplement quality assurance scheme data with information on risk factors for STM infection.Data were collected from Quality Assurance (QA) schemes, to review and analyse the practicality and relevance of using this information for risk factor analysis. An analysis of the data against corresponding ZAP sample results showed that the quality of the data was not sufficient to allow the use of many of the variables collected and the difference in data collection between the three schemes meant that many variables also had to be dropped to allow all schemes to be used. However, the following variables from the QA schemes could be used:- the use of flooring types in finisher houses; whether wet or compound feed fed to finishers; whether home mix feed used; whether all pigs were kept indoors and the number of sows and finishers. A literature review and also expert opinion was used to identify factors important to the prevalence of Salmonella in pigs and to decide upon the optimal length of the supplementary questionnaire, to generate a form that would not overburden participating farmers but would collect all those significant factors not already supplied by the QA schemes and any other sources. Previous studies were used for ideas on how to word the questions successfully and the form was piloted with five pig farm staff members, to look at the ease of completion. Copies of the questionnaire are available from CERA, VLA Weybridge and detailed analysis will be reported in due course through Defra project OD0215.

1.5 Combine MJ ELISA data, QA data and questionnaire data for epidemiological analysis.The data routinely collected by the three QA schemes and the ZAP sample data were combined using a combination of quality assurance membership identifiers, and also postcodes and herdmarks where available, and several data validation steps were added to ensure that samples were not joined to more than one holding. By combining and merging the datasets, a population of over 1,500 pig holdings was generated with an average of 30 samples per farm. The data were analysed to identify risk and protective factors for Salmonella infection (defined as positive or negative by a 0.25 cut-off value), and to indicate the types of analyses that were possible with this type of data. Only two variables were found to be significant in the multivariable model, with samples from farms in Yorkshire and the Humber and those that used solid flooring for finishers significantly more likely to be positive. Univariable analysis also indicated significant regional differences in pig management and spatial analysis showed that high prevalence holdings in Yorkshire and the Humber were more clustered in space than low prevalence holdings. The use of the data provided a large representative population for analysis, which detected even weak associations between variables, however the data quality of some of the variables was poor

SID 5 (Rev. 3/06) of 30

and many variables had to be dropped from the final dataset. Recording errors in the identifiers caused problems in matching samples to holdings.

During June 2007 and October 2008, supplementary forms were collected and were linked to an update of the QA scheme data, monthly regional summaries of meteorological data and up to four years of ZAP records. The final dataset consisted of 566 holdings with a mean of 225 ZAP samples per holding.

One member of the VLA project team registered for a part time PhD at the University of Liverpool and the tasks listed in 1.4 and 1.5 are contributing to this training.This final report, summarising the epidemiological components under objective 1 of this project, satisfies project milestone 01/03.

One final activity under objective 1 was to submit a proposal for a case-control study to the Food Standards Agency, to estimate the population attributable fraction (PAF) for human STM disease originating from UK pigs (see secondary project milestone S01/04). This proposal was rejected. However, recently a new proposal has been developed in collaboration with the Health Protection Agency (HPA) and London School of Hygiene & Tropical Medicine, which may lead to implementation of this research at some later date. It is possible that there may be some prejudice against funding for research into the human impact of food-borne zoonoses across the whole food chain, as no single Government Department or Agency considers these conditions to lie exclusively in their domain and thus to merit the dedication of limited resources when compared to other priorities.

MICROBIOLOGY – OBJECTIVE 2The aim of this work package was to provide prevalence data, semi-quantitative estimates of Salmonella burden and molecular evidence of transmission routes, applying newly developed techniques to address key data gaps in the microbial risk assessment. The role of Salmonella originating from pigs in human infection was further elucidated.

2.1 Apply appropriate strain typing methods to selected STM isolates of known provenance in order to: Examine persistence of strains on farms and through abattoirs Compare human and pig isolates Investigate transmission between and within pens

Examine persistence of strains on farms and through abattoirsThe details of this study are given in appendix 1 and will soon be available at http://www.foodbase.org.uk (FSA Final report MO1040, objective 5 - Use of strain typing methods to investigate persistence of S. Typhimurium clones from farm to abattoir and to investigate whether such methods may be useful for attribution of human infection). This report includes the studies undertaken in OZ0323 and previously reported to Defra (milestone 02/01). Briefly, four farms (0051B, 1652Z, 1922Z and 1925Z) studied in project OZ0316 were selected for further investigation based on the isolation of Salmonella spp. on more than one sampling occasion on each farm and isolation of Salmonella spp. from pigs at the abattoir. On each farm, a batch of pigs was sampled longitudinally through the production cycle to the abattoir. Samples were taken from the environment and pooled faeces. Summary reports on salmonella status were prepared for each participating abattoir and for individual farms (secondary milestone S02/01). At the start of the study, participating farms were sampled pre-trial, after the previous batch of pigs had been removed. They were then sampled post cleaning, then at 3 days following introduction of the pigs to the farm, and again at 4, 8, 12 and 16 weeks after introduction. Prior to the pigs leaving the farm, faecal samples were collected from each pig. At the abattoir, samples were collected from the lorry in which the pigs were transported, lairage pens before and after the study pigs were introduced, rectal and carcase swabs from the study pigs following slaughter and caecal contents. Using standard isolation, serotyping and ‘phage typing methods for Salmonella (VLA SOPs BAC212, CBU001, 005, 042, 044, 047, 063, 075, 092, and 0111 and CBU0146), 155 S. Typhimurium isolates were isolated and identified from farm 0051B, 27 from farm 1652Z, 53 from farm 1922Z and 32 from farm 1925Z. All S. Typhimurium isolates were subjected to molecular characterisation using Pulsed-Field Gel Electrophoresis (PFGE) and Variable Number Tandem Repeat (VNTR) analysis. Three of these farms were studied within FSA project (M01040) and isolates from farm 0051B were characterised within Defra project OZ0323. Results from the latter farm showed S. Typhimurium was collected at most stages of the study except during visits 4 and 5 where the only Salmonella serotypes detected were London, Manhattan, Panama and Reading. Seventeen different VNTR types were identified among the isolates from this farm, with many only differing by a single allele or repeat. By PFGE, 5 different XbaI profiles were identified among representative isolates of each VNTR type. ‘Phage-typing also identified five types, DT104, U288, U302, DT193 and 170b. The most prevalent isolate on this farm at the start of sampling had allele string 2-9-5-0-0. This type was present at the first sampling visit to this farm before cleaning. Following introduction of pigs isolates with allele string 2-9-5-0-3 were detected. After the 8 week visit VNTR type 4-15-4-0-2 became the most prevalent isolate on the farm and these remained until the end of the study. From ‘phage-typing information it appears that isolates of type U288 became most prevalent after the week 8 visit. Prior to this isolates of phage-type 170b were most prevalent on this farm.

SID 5 (Rev. 3/06) of 30

Key findings from the four farms included specific VNTR types that persisted on the farm and were also detected at the abattoir and on the final carcass; the suggestion of acquisition of S. Typhimurium and S. Kedougou isolates in the lairage pens which were not previously seen on the farm (the S. Typhimurium isolates were able to spread to the mediastinal lymph nodes and caecal contents following acquisition); S. Typhimurium, present in the pigs before slaughter being eliminated or reduced below the levels of detection during processing so that it was not detected on the final carcass. Molecular methods used for typing of isolates during this study have been valuable in studying infection dynamics within batches of pigs. They have allowed identification of points in the production process where Salmonella may have been introduced for further investigation. These techniques have been useful for tracking strains through the production cycle and have identified changes in prevalent clones on farms, which was not possible with serotyping information alone.

A poster presentation was made on this work (secondary milestone S02/03):Kirchner, M., Clifton-Hadley, F.A., Miller, A., Snow, L., Davies, R.H. and Cook, A.J.C. Applying VNTR and PFGE for tracing of Salmonella Typhimurium on a pig farm. 8th International Meeting on Microbial Epidemiological Markers, Zakopane, Poland, 14-17 May 2008.

A draft publication on this work has also been written entitled:Use of molecular typing methods to follow S. Typhimurium during the production cycle of pigs reared on four farms in the UK.

Compare human and pig isolatesThe aim of this study was to assess whether molecular characteristics could be identified in a panel of S. Typhimurium isolates from different sources (Table 3) which could be used to attribute an isolate to a source. It was thought that this may provide evidence on whether or not locally produced pork is a significant source of S. Typhimurium for the human population in the UK. Details of the study are given in appendix 2 and will soon be available at http://www.foodbase.org.uk (FSA final report on MO1040 Objective part B - Study the trends in pig related STM ‘phage-types, e.g. U288, U310, U308a, DT208, DT193 and a ubiquitous ‘phage-type) which documented jointly-funded work. The work has previously been presented to Defra (milestone 01/03). S. Typhimurium isolates collected between December 2006 and November 2007 from animals and humans were subjected to molecular characterisation by PFGE and VNTR (Table 3). Isolates were selected based on ‘phage-type to include the pig-related ‘phage-types (U288, U308a, U310, DT193 and DT208) and DT104 and U302 were selected for comparison. 784 isolates were characterised. It was found that VNTR gave better discrimination than PFGE for all ‘phage-types studied. This was confirmed by Hunter and Gaston discrimination analysis comparing PFGE, VNTR, ‘phage-type and source. Multiple correspondence analysis using PFGE or VNTR data found that cattle and poultry isolates clustered together. Pig isolates clustered along with U288 ‘phage-type and humans were most closely associated with DT193 and U302 ‘phage-types. Logistic discriminant analysis was used to determine which parameters were best at assigning an isolate to source. It was found that a combination of ‘phage-typing, PFGE and VNTR gave the best determination of source for pig, human and cattle isolates. Forty-three percent of the isolates tested were from pigs. VNTR alleles 4-15-4-0-2 were most frequently detected in pigs. The most frequent VNTR alleles in pig DT193 and U288 were found to be very similar. Two-hundred and four human isolates of ‘phage-types U288, DT104, DT193 and U302 were analysed. For the’ phage-types DT193 and U288 it was found that the most frequent VNTR alleles were similar for pig and human U288 isolates but different for human and pig DT193 isolates. Ninety-three isolates from turkey and chicken were also analysed. Most of the isolates tested were of’ ‘phage-type DT104 and other non-pig related ‘phage-types. Only two isolates of DT193 were collected during the study period. Of interest is the identification of a large cluster of isolates from multiple turkey farms with identical PFGE and VNTR profiles. A small number of food isolates was obtained from the FSA meat survey and also pig carcass swab samples were tested from the EU baseline survey for Salmonella in slaughter pigs, project FZ2023. There was a wide variation in the VNTR and PFGE profiles of the isolates tested. Some isolates were identical by PFGE and VNTR to S. Typhimurium from pigs. Cluster analysis of VNTR results demonstrated that there were 4 large clusters. In general, these clusters were distinguishable by ‘phage-type and source. Twelve smaller clusters were identified, which contained isolates from multiple sources that were identical by VNTR and PFGE. These included types which were detected in humans and pigs, humans, pigs and food, and pigs and carcass swabs (this latter might be expected due to contamination of pig carcasses with pig-associated isolates).

SID 5 (Rev. 3/06) of 30

Table 3. The number of isolates selected for testing from each species and ‘phage-type. Species ‘Phage-type

DT104 DT193 U288 U302 DT104b Other RDNC UNTY Not testedHuman 65 97 26 16 - - - - -Pig 17 110 171 32 5 6 1 2 33Chicken 5 2 - - - 1 3 - 24Turkey 45 2 - 11 2 6 - - 15Avian other 1 1 1 - - - - - 1Cattle 33 9 - 6 7 - - - -Sheep 6 - - - - - - - -Horse 2 2 1 - - - - - -Pets 4 - 1 - 1 - - - -Environmental 2 1 2 - 1 - - - -Misc animal 3 - - - - - - - -

Pig carcass swab 1 8 10 2 3 4 - 6 1

FSA_MEAT 2 4 - - - 4 - 4 -

RDNC = reacts does not conform; UNTY = untypable

Investigate transmission between and within pensData that were collected during a longitudinal intervention study in a previous project (OZ0316) were analysed to investigate transmission between and within pens of pigs. As described in the previous microbiology section (“Examine persistence of strains on farms and through abattoirs”), VNTR typing was used to describe the movement of Salmonella between pens on four of these farms. The data from all 48 farms were used to populate a transmission model, as described in the modelling section (Objective 4 - Estimate parameters for between and within pen spread for the farm module of the MRA) later in this report. The results were used to update the MRA, as described under objective 3 below. A draft report on application of appropriate strain typing methods was completed to satisfy project milestone 02/01.

2.2 Use semi-quantitative methods to estimate infectious burden in the farm and abattoir environment, in individual pigs and on carcassesPigs from 12 farms that had been recruited to a previous study were followed from farm to abattoir to investigate the burden of infection within individual pigs or on their carcasses after slaughter.

The samples were subjected to semi-quantitative analysis. A total of 346 samples were taken, this included rectal faeces from individuals before leaving the farm (n=52) and then samples from the mediastinal lymph nodes (n=95), caecal contents (n=178) and carcass swabs at slaughter (n=21). Of the positive samples, 67% were STM, followed by 25% S Derby; a further 8 different serotypes were found in less than 2% of samples each.

In order to provide a semi-quantitative estimate of the burden of infection, a dilution enrichment technique was applied to positive samples. This involved preparing serial tenfold dilutions of the original sample material in buffered peptone water and following up with a sensitive isolation technique. An indication of the level of Salmonella present was obtained from calculating back from the highest Salmonella positive dilution.

Figure 1 (page 15) shows that for the faecal, caecal and lymph node samples, Salmonellas were still detectable at a dilution of 1:10000 of the original sample, even though only approximately 2% of samples were still positive at this stage. This last dilution would be approximately equivalent to 103 to 104 colony forming units per gram of original sample. The lymph node samples appear to show a slower drop in the proportion positive with each dilution; however, all three show a similar rate of decline. In comparison, of the 21 carcass swabs, only one was positive after the initial 1:10 dilution suggesting the burden of infection in carcass swabs is significantly lower (only about 1-10 units per swab) than when samples indicative of internal infection were collected.A draft report was prepared on this work to satisfy project milestone 02/02.

SID 5 (Rev. 3/06) of 30

0

20

40

60

80

100

120

Original (100ml) sub-sample (10ml) 1:10 1:100 1:1000 1:10000

Dilution of original sample

% p

ositi

ve s

ampl

es

faecal per rectumCaecalMLNcarcass swab

Figure 1: Proportion of positive samples by type of sample for stage of testing.

FARM TO FORK MICROBIAL RISK ASSESSMENT (MRA) – OBJECTIVE 3

3.1 Review scientific and other literature published since July 2003 and update MRA accordinglyA review of recent literature on pig Salmonella uncovered four main pathways for control - either to stop Salmonella from being introduced onto the farm; eradicating Salmonella once it has infected the herd; stopping Salmonella spreading within the herd; and improving the herds chance of resisting infection e.g. by vaccination, and each of these measures are discussed in the review. The most effective and economic measures to reduce Salmonella infection were reportedly:-

To limit the risk of Salmonella entering the herd. This would include the monitoring and heat treatment of feed entering the farm; increased biosecurity to reduce the access of pets and wildlife to the pigs and their buildings; and improved biosecurity for human visitors (e.g. pen specific overalls and footwear, along with boot dips). Access to certified Salmonella-free pigs through Accreditation schemes would allow farmers to ensure that they were importing pigs from herds with the same or a higher health status; however, there is no such scheme in the UK at present.

To control the spread of infection by using an all-in/ all-out system for pig buildings with slatted floors, combined with stringent cleaning and disinfection between batches, pen separation to stop contact between different pig groups, minimising mixing and isolating sick animals would reduce risk of faecal-oral transmission;

The use of meal feed, wet feed or fermented liquid feed would help control the levels of Salmonella in the gut by promoting a reduced gut pH, which is less favourable for growth of Salmonella. However, meal or wet feed reduces pig growth rates and fermented liquid feed requires specialised equipment. Organic acids supplementation via feed or water has also been advocated, although the efficacy in individual farm trials in UK has been limited. The use of live vaccines, the reduction of respiratory diseases and pig stress (e.g. by less transporting, less handling) would also improve the control of Salmonella infection and shedding in pigs.

It is acknowledged that there is no universal solution that is applicable to all pig husbandry systems.

The farm simulation model was incorporated into a farm-to-fork microbial risk assessment (MRA) for Salmonella in pork, developed as part of project OZ0316. The MRA incorporates 6 modules namely, farm, abattoir, further processing, distribution, retail and consumption and has been revised and updated with data obtained from a review of the scientific literature, expert opinion and results from current research. We used data collected by ABP (2007) and GQA (2007) to model the variability in herd sizes on UK farms. A first order inactivation kinetics model is used to estimate the reduction in Salmonella during cooking, where an exponential decay curve is fitted to appropriate data to predict the appropriate D value for different temperatures. We used preliminary results

SID 5 (Rev. 3/06) of 30

from objective 4 of the current report to estimate the within pen farm transmission parameter. Data from objective 1 are used to help validate the farm module.

3.2 Develop a stochastic breeder-weaner model for STM infection and incorporate it into the MRAIt is important to understand the transmission dynamics of Salmonella infection within pig herds in order to implement effective on-farm control measures. Therefore, a stochastic simulation model was developed to investigate the transmission dynamics on a managed pig farm, including both breeder sows and finisher pigs. Based on available data and model assumptions, it was deduced that endemic infection would occur in 66% of clean farms where infection was introduced via one pig. Approximately 26% (5th and 95th percentiles of 0% and 75%) of pigs sent to slaughter from these farms were estimated to be carriers and 2.3% (5 th and 95th percentiles of 0% and 14.29%) were estimated to be excreting Salmonella in their faeces. Figure 2 shows the distribution, between farms, of the prevalence of excretors among the different groups of pigs, 40 weeks after the initial seeding of infection. It can be seen that there is not a huge variation between groups, with few farms having a prevalence of over 0.3. A draft report on the stochastic breeder-weaner model was circulated to satisfy project milestone 03/01.

Figure 2: Distribution over farms of average prevalence of excretors among pig groups 40 weeks after initial infection. 3.3 Evaluate other methodological approaches for components of the MRA especially where data are sparse, e.g. Bayesian inference and modify the MRA as appropriateIn the updating process different methodological approaches have also been applied to parts of the MRA to make better use of the available data. Methods from systems using delay differential equations have been adapted for use in the stochastic MRA and used to model the variation in duration of shedding of excretor pigs. The previous farm model has been modified to incorporate the possibility of re-infection (pigs in the carrier state, which do not currently excrete Salmonella in their faeces, may move back to the excreting state). We have adopted a method used by other risk assessments (e.g. Nauta et al. 2001, Kosmider et al. 2008) to model the predicted number of Salmonella organisms on each joint from a contaminated pig carcass, taking account of the fact that Salmonella are likely to be clustered to specific areas of the carcass. We have used Bayesian analysis methods to update the previous data for temperature of products during home refrigeration with data from a more recent study. This method allows us to define a new distribution for the variation in temperature using the data from both studies. We revised the cross-contamination module by including more data from a recent study and incorporating more of the natural variability associated with the parameters in the model. The MRA aims to estimate the risk to humans of Salmonella from consuming pork products.

Table 4: Average risk of infection from chops, bacon and sausages, associated confidence intervals and percentage of contaminated products

Product Average risk

95% confidence interval

% of contaminated portions (AH)

% of contaminated portions

No. servings consumed per person

Average number of human cases/ year

SID 5 (Rev. 3/06) of 30

causing illness per yearChops

1.56E-071.45e-007 , 1.68e-007 1.37E-06

11.43 13 96.5

Bacon1.42E-09

1.10e-009 , 1.75e-009 1.09E-07

1.32 26 1.7

Sausage4.06E-07

4.00e-007 ,4.12e-007 1.76E-05

2.31 26 502.4

Table 4 shows that the MRA estimates that the average risk from pork chops, bacon and sausages are 1.5614e-007, 1.4224e-009 and 4.0564e-007, respectively, therefore suggesting that sausages are the more risky product, although the risk from all products is low. The model estimates that there is wide variation in the risk from individual portions of the product types, with some portions more likely to cause illness than others due to a larger concentration of Salmonella.

ReferencesABP (2007). Approved British pigs assurance scheme.GQA (2007). Genesis quality assurance pig scheme.

Kosmider, R., Nally, P., Simons, R., Brouwer, A., Cheung, S., Snary, E. & Wooldridge, M. (2008). Attribution of human VTEC O157 infection from meat products: a quantitative risk assessment approach. Submitted Nov 2008, Risk Analysis special edition

Nauta, M. J., Evers, E. G., Takumi, K. and Havelaar, A. H. (2001) Risk assessment of Shiga-toxin producing Escherichia coli O157 in steak tartare in the Netherlands.

MODELLING STM INFECTION – OBJECTIVE 44.1 Estimate parameters for between and within pen spread for the farm module of the MRA The aim of this section of the project was to develop a methodology, which could be applied to longitudinal pooled sample data from a number of farms, collected as part of project OZ0323, to estimate transmission coefficients of the spread of Salmonella infection between and within pigs housed in pens within a finisher house.

Pooled sample data, collected from 171 finisher houses on 48 farms randomly selected from a national database, were analysed as part of this section. For each house, pens were monitored from entry of pigs at approximately 10 weeks of age up age of slaughter at approximately 24 weeks of age. In general, one pooled sample was taken per pen sampled, per time point.

In order to estimate between and within pen transmission rates, an extension of the classical susceptible-infected-recovered model was employed, with the addition of serological response and the possibility of pigs reverting back to shedding Salmonella or losing immunity to re-infection. A starting prevalence was estimated from the pooled sample data for each pen sampled and simulations of Salmonella transmission were performed for given within and between-pen transmission rates, using the sample prevalence as an initial value. Maximum likelihood methods were then used to determine the most likely transmission rates for the data. The model was validated by using it to estimate transmission parameters from data that has been simulated using known transmission values.

The most likely transmission parameters, given the data, had means of 0.053 for between pen transmission and 0.064 for within pen transmission. Approximately 7% of parameter combinations were significantly similar to the combination that corresponded to the maximum likelihood; within these, between-pen transmission ranged from 0.039 to 0.1 and within-pen transmission ranged from 0.05 to 0.134. These values are of the same magnitude obtained by Hill et al in 2007.

The effect of sample size and frequency of sampling on the ability of the model to predict transmission parameters from the data was investigated. Results showed that with low numbers of iterations, the model can correctly predict transmission rates if the number of samples taken per pen per time point is large enough. The results obtained suggest that the model is more sensitive to sample size than the frequency of sampling, implying that if the sample size is large enough, the frequency of visits can be reduced without significantly affecting the ability of the model to predict within and between pen transmission rates. The lack of power in the data from low numbers of samples being taken is overcome by the large number of houses included in the dataset.

The results imply that there is little difference between within and between-pen transmission rates. Despite this being unexpected, the results are in-line with previous studies. In order to investigate this further, a higher number of samples, from more pens within the same house, should be taken. We recommend using the model in experimental design to determine the sample size required and the frequency of sampling. The model can only predict transmission parameters up to the (limited) resolution of the grid used to draw the candidate parameter values, this could be improved upon by coupling the model with an MCMC approach to enable maximum likelihood estimates to be obtained with greater resolution.

SID 5 (Rev. 3/06) of 30

4.2 Develop an approach to estimate herd-level prevalence from pooled samples.The objective of this study was to estimate sample sizes for pooled-pen faecal sampling to detect Salmonella in pig herds. Previous work (Arnold et al., 2005) has looked at the sensitivity of pooled faecal samples, but there was uncertainty about how applicable these results were to pooled samples taken in the field due to the artificial nature of the creation of the pools. Therefore, the results of parallel sampling of pooled and individual samples from several pens of pigs from nine farms were analysed using Bayesian methods to estimate sensitivity in the absence of a gold standard to obtain a comparison between pooling and individual sampling. Furthermore, an estimate of the degree of pen-level clustering of infection was obtained, as this impacts on the required sample sizes. The resulting estimates of the sensitivity and the pen-level clustering were then used to estimate required sample sizes for a range of prevalence.

Data collection and sample cultureThe nine farms were amongst 48 that had been selected at random from a national database and recruited to another study (OZ0316). The data used in this study were from the final, pre-slaughter visit, at which point replicate pooled samples were collected and additional individual samples were obtained by sampling per rectum. All samples were cultured for the isolation of Salmonella bacteria using standard methods.

Statistical methodsThe data in the present study consist of a set of herds where both pooled faecal samples and individual rectal samples are taken from selected pens. This is a parallel situation to one considered in Branscum et al. (2004, 2006) where a Bayesian approach was developed to estimate both the region-level prevalence for a set of regions and the within-herd prevalence of each herd within each region. Therefore, we adopt a similar approach to that developed in Branscum et al. (2004, 2006)], where in the present study herd is the equivalent of region and pen equivalent to herd in Branscum et al. (2004, 2006). One additional enhancement of the Branscum model was necessary, since the proportion of pig pens infected will be proportional to the number of individual pigs infected. The relationship between the animal level and pen level prevalence was not known, but was estimated using data from M.E. Arnold et al. (2009) – See Figure 3.

Figure 3. How the proportion of pens infected varies according to the animal-level prevalence for Salmonella in pigs. The observed proportion of positive pens for the positive farms in a UK study are given by crosses (the observed values are lower than the predicted proportion of positive pens at low prevalence since it is likely that some truly positive pens will be false-negatives at low prevalence).

The results of the sampling of individual-level and pooled samples are given in Table 5, along with the estimated individual-pig prevalence from the Bayesian model. Table 5 shows the increased sensitivity of pooled sampling compared to individual-level sampling.Table 5 - The farm-level data for the results of the pooled and individual sample testing for Salmonella in pigs in the UK, and the estimated mean prevalence of animal-level infection for each farm.

Farm Pooled faecal sample

Individual faecal sample

Number of pens positive/number of pens tested

Median herd prevalence (95% credibility intervals)

1 31/40 10/43 7/8 0.53 (0.30, 0.75)2 0/15 0/46 0/3 0.06 (0.00, 0.35)3 1/20 - 1/10 0.09 (0.02, 0.26)4 17/25 - 6/8 0.49 (0.25, 0.71)

SID 5 (Rev. 3/06) of 30

5 0/20 0/35 0/7 0.00 (0.00,0.20)6 0/25 0/32 0/5 0.02 (0.00, 0.24)7 6/35 6/45 2/7 0.18 (0.05, 0.40)8 6/25 0/45 4/5 0.28 (0.11, 0.50)9 3/25 3/41 1/5 0.20 (0.06, 0.45)

Sample size estimationThe number of samples required for a detection probability of 95% for animal-level prevalence on farm of 30, 20, 10, 5, 2 and 1% was estimated by simulating the prevalence in each pen and calculating the resulting probability of detection for a range of sample sizes. The number of samples required was taken to equal the least sample size for which the median probability of detection was greater than 95%, sampled over 10,000 iterations. To explore the impact of the number of pens on the number of samples required, 3 scenarios were considered: 200 pens (representing the case where the number of pens sampled is much larger than the number of pens in the farm), 20 pens and 10 pens. Results are given in Table 6, and show the impact of low prevalence on the number of samples required. As the number of pens reduces the number of samples also reduces, as also occurs in the case of individual samples..Table 6 - Estimated number of pooled faecal samples to be collected to detect at least one positive sample in a pig herd with 95% certainty over a range of Salmonella prevalence.

Number of pens in herd

Percentage of infected pigs in herd30 20 10 5 2 1

200 6 9 19 38 90 15020 6 9 16 24 58 11010 6 8 11 22 54 105

In conclusion, this study has estimated sample sizes for the number of pooled samples for detection of Salmonella in pigs. It has also provided an estimate of the degree of clustering of infection at pen-level, which influences the number of samples required for collection, and it would be helpful if more data could be collected on this. It has also provided a method for estimating individual and pen-level prevalence of Salmonella in pig herds. This work will form the basis for the analysis of within-herd data from the current EU baseline survey for Salmonella in breeding pigs.The results of the estimation of between and within pen prevalence were reported to the project team, satisfying project milestone 04/01.

ReferencesArnold, ME, Cook AJC, Davies RH. A modelling approach to estimate the sensitivity of pooled faecal samples for isolation of salmonella in pigs. Journal of the Royal Society Interface 2005; 2: 365-372.Arnold ME, Cook AJC. Estimation of sample sizes for pooled faecal sampling for detection of Salmonella in pigs. Accepted for publish – Epidemiology & InfectionBranscum AJ, Gardner, IA, Johnson, WO. Bayesian modelling of animal- and herd-level prevalences. Preventive Veterinary Medicine; 2004; 66:101-112 Branscum, AJ, Johnson, WO, Gardner, IA. Sample size calculations for disease freedom and prevalence estimation surveys. Stat Med 2006; 25: 2658-2674.

SPATIO-TEMPORAL ANALYSIS OF ZAP DATA – OBJECTIVE 5

5.1 Assess spatial and temporal variation in the ZAP database and examine covariate relationshipsRoutine Quality Assurance (QA) and supplementary questionnaire data (see section on questionnaire development under objective 1 above) were modelled against the ZAP sample records via two main methods – a logistic multivariable model, with the holding included as a random effect and using the 0.25 cut-off point for each ZAP sample ratio; and a linear multivariable model, also using farm as a random effect but modelling the sample ratio result directly. Spatial and temporal terms were added to the model and also explored individually via graphs and GIS maps. Marked point process analysis was not appropriate for this type of study, although geo-statistical approaches were used to account for the covariates identified in the final linear multivariable model when analysing the spatial correlation of the ZAP ratio results.

The results of these models indicated a number of variables that were significantly associated with Salmonella infection, and both models agreed that the variation caused by sampling from abattoirs needed to be considered. The models identified variables linked to respiratory and wasting health problems, and previous diagnoses of clinical salmonellosis, as risks. A conventional farm type (i.e. indoor breeder-finisher unit) was found to be protective, as were certain feed-related variables, e.g. the use of home mixed rations, an increased percentage of barley concentrate in feed or use of fermented liquid feed. Increased numbers of visitors to holdings also

SID 5 (Rev. 3/06) of 30

increased risks, such as increased number of pig deliveries, school/ student visits, dead stock collections and slurry/bedding collections. Conversely, holdings that did not have a contracted vermin control programme were more at risk. Farms that were visited by pig specialists (nutritionists, breeding experts) were at a reduced risk. The presence of camelids and dogs on the holding was shown to have an association with Salmonella although these animals were only present on few holdings and so these results may have arisen by chance.

All-in/all-out batch production on a shed basis was associated with an increased risk of Salmonella infection. However, this paradoxical finding may be because these buildings belong to specialist contract finishers and are re-stocked with pigs from multiple sources. In contrast, finisher pig accommodation that is continuously occupied may be more frequently associated with breeder-finisher enterprises. The results emphasise that all-in/ all-out management alone may not be sufficient to control Salmonella infection in pigs.

In the linear model, both yearly and quarterly cycles were found to be significant, with the yearly cycle also included in the logistic model. Large differences to long term averages in the hours of sunshine, a high mean temperature and high actual rainfall were also identified as risk factors in the linear model, whereas anomalous rainfall was significant in the logistic model. Temperature has been shown previously to be associated with pig stress and with Salmonella prevalence.

The two models yielded concordant results with respect to most of the variables identified, and those that differed may have come from artificially creating a cut-off point for the binary outcome. Variables restricted to only the linear model may be more related to samples with very high optical density results whereas those only in the logistic model may be linked to samples with an average ratio only just above the 0.25 cut-off.

Both final models selected a high farm density at the 10 km radius as a risk and also showed that samples taken from farms in Scotland were less likely to be positive. When the covariates detected in the linear model were added to variograms, the residual spatial correlation shown by earlier studies was not discernable. This result may pinpoint that once all co-variants are accounted for, these two variables were the only significant spatial factors. The results from this research are being prepared for submission to peer-reviewed journals.

5.2 Modelling the probability of a premises changing infection status between visitsThe time-dependent probabilities of moving from negative to positive were investigated by analysing the ZAP data via two models – a null model with a common transition probability, which is the same irrespective of the time of year; and a saturated model with different transition probabilities at each time point. Then a variety of sub-models, in between the two models, were considered and compared using likelihood ratio tests e.g. a model to examine 3 month periods (Jan-march etc) irrespective of the year, consistent with some kind of seasonal effect. The first step was to incorporate all of England and Wales as a single unit (so ignoring potential geographical heterogeneity at first) and the saturated model provided a superior fit over the null model and the sub-models. Then the patterns in three separate regions were studied. East of England matched the national results with the saturated model providing the best fit, whereas in Yorkshire and the Humber a sub-model with a separate transition probability between quarters was the most suitable. In the South West, the saturated model offered no significant improvement over a common mean over time and the sub-model offered only a marginally significant improvement over the null model. The South West region had the smallest number of farms of these three regions, and in the presence of more data a seasonal effect might be formally demonstrated. The probability of going from negative to positive at any time point is appreciably higher in the East of England, Yorkshire and Humberside than in the South West. It is potentially interesting that these higher probabilities occur in higher density regions.A draft paper based on these results was prepared, satisfying project milestone 05/01.

5.3 Use of MJ ELISA data to predict areas with an increasing risk of salmonella infectionThe K-function and approaches from the field of geostatistics (variograms) were used to study routine data from the ZAP scheme. The more powerful variogram-based analysis, which modelled the prevalence data directly, did provide preliminary evidence of residual spatial structure in one of three regions studied (East England), whereas the K-function-based analysis, which relied on outcomes determined by arbitrary cut-off values, did not. This analysis highlighted some spatial dependence over shorter distances, up to around 20km. In contrast, for Yorkshire and Humber and the South West, neither the variogram nor K-function based assessment suggested any evidence of spatial dependence. The physical mechanisms underlying this structure remain unclear: spatial structure might be present as a result of shared spatially structured (second-order) or non-spatially structured (first-order) risk factors, transmission processes, or a combination of both. Information concerning the presence of geographically localised regions at higher risk of Salmonella infection can be used in the design of targeted surveillance and control strategies. Furthermore, it can begin to indicate underlying determinants of infection that could be targeted in such control programmes.

In an effort to understand how prevalence varies throughout the year at the farm-level, the kernel smoothed point estimates of prevalence within farms over time for the whole of England and Wales were plotted, using a Normal

SID 5 (Rev. 3/06) of 30

kernel and a bandwidth of 100 days to highlight broad seasonal patterns rather than more localised variation. The best estimates of herd prevalence were also plotted across the time period. There was some evidence that within-farm prevalence is highest in both years studied from September to January and at its lowest in June to August. In all regions, to a greater or lesser degree, the characteristic peak in winter and dip in summer is observed.

A Generalised Linear Mixed Effects model was fitted to the prevalence data from each region separately, so that relationships between prevalence and explanatory variables could be studied and the models examined for similarities and/or discrepancies between the regions. This approach was used to incorporate cycles (reflective of seasonal patterns) and also trend (to see whether Salmonella levels have systematically increased or decreased over time). In all regions, the model highlights some evidence of both 6 and 12 month cycles. Specifically, in all cases there is a suggestion of a primary peak in prevalence within farms in Autumn (September/October), a secondary, smaller peak in January/February, and a definite decline around June and July. Evidence concerning trend differs in the three regions; there is no evidence of a trend in Yorkshire and Humber, evidence of a marginally significant increase in East of England and evidence of a marginally significant decrease in the South West. These results are reported in a paper which was accepted for publication by Preventive Veterinary Medicine in January 2009.

Data from a previous study (OZ0316) had been provided to a MSc student at Lancaster University, who gained a distinction for her thesis. That work was expanded by modelling logged MJ ELISA signal:positive quantitative data directly. Seasonality (12 and 6 month cycles) and a small number of covariates for two areas of the UK (East of England and Yorkshire and Humberside) were added to study spatial and temporal trends. A novel mode fitting approach utilising an EM algorithm was used, as the data set was large and included repeated measures over time. The results provide evidence of the effects of season, specialist breeder/breeder-finisher status and location on Salmonella levels. A key advantage of the model over simpler fixed effects models or threshold exceedance criteria is that it enables the identification of the relative contribution of different sources of variation in Salmonella levels. This knowledge is important for the design of targeted control strategies and sampling schemes. Within-animal across-time variation is large by comparison with farm level non-spatial and spatial variation, suggesting that the repeated sampling within farms which is currently implemented as part of ZAP is warranted.

The advantage of explicitly modelling spatial dependence is the ability to identify local hotspots, being areas within which farms appear to have intrinsically higher levels, having controlled for farm specific covariates and seasonal variation. A limitation of this work is that it does not fully respect the temporal effects – as spatial field and farm-level random effect are assumed to be constant over time. A draft report describing this work was circulated to the project team, satisfying project milestone 05/02. The results will also be submitted for publication in a peer-reviewed journal.

SOCIO-ECONOMICS – OBJECTIVE 66.1 Develop least-cost models for interventions at farm and abattoir level within the MRAThis section presents the cost calculation of interventions to control Salmonella in the British pig industry. Costs are calculated for farm and abattoir level interventions. Preliminary research identified biosecurity, rodent control, feeding with organic acids and vaccination as the most effective interventions at the farm level (OZO316). At the abattoir several different interventions have been scrutinized in the literature namely vertical scalding, bunging, washing and drying and second singeing (Richmond, 1991; ICMSF, 1998; Wong et al., 2002; Bolton et al., 2002b; Goldbach and Alban (2006); McNamara et al. 2007 James et al. 2008). However here we only consider the costs of rodent control, organic acids feeding, vaccination, vertical scalding, bunging and carcass washing and drying.

To enable comparisons between interventions a decision has to be made in terms of a common basis for cost calculations. There are two possibilities: perform all calculations on the basis of breeding sows or of slaughtered pigs. The former is more appropriate for calculations at the farm level as sows are the basic production unit, also there are accurate annual statistics on the total number of breeding sows in the UK. Using slaughtered pigs enables an immediate comparison between farms and abattoirs. Once the basis of comparison is decided the parameters required to determine each intervention’s costs must be specified.

There is no previous research in the UK focusing on the cost of bio-security on pig farms, which limits a precise determination of costs. A number of parameters and assumptions had to be considered, and these were based on previous studies carried out in Denmark (Goldbach and Alban 2006) and the United States (McNamara et al 2007) along with data from the UK pig industry (BPEx 2007) and DEFRA (2007). For each intervention a baseline model was calculated, which was then used in the sensitivity analysis performed subsequently. The results from the economic analysis were used to estimate the costs of STM reduction interventions, which satisfied project milestone 06/01.

6.2 Investigate economic optimisation of interventions along the food chain to reduce human STM illnessTo calculate a cost-effectiveness ratio for each intervention we combined the estimated economic cost of each of the interventions with the predicted effectiveness estimated by the MRA. In order to jointly consider both the farm

SID 5 (Rev. 3/06) of 30

and abattoir interventions used comparable cost and effectiveness measures for all interventions. Therefore, the cost measure we used is the industry cost in terms of pigs and the effectiveness measure is the average percentage change in concentration of Salmonella on the pig carcass from the baseline model immediately prior to chilling at the abattoir. These measures are chosen as they are applicable to all interventions. While we acknowledge that the percentage change in prevalence, rather than concentration, would be more applicable to the farm interventions, it would be far less applicable to the abattoir interventions.

The cost-effectiveness ratio for intervention i, E(i), is calculated by dividing the percentage change in carcass contamination, P(i), by the cost associated with that intervention, E(i). We also multiply by a factor of 10,000 to make the results clearer.

Our findings suggest that there may be opportunities for savings if interventions are concentrated in abattoirs. However, this would contradict most common practice and previous studies, which recommend multiple barriers for effective pathogen reduction. Since our study was initially only focused at farm level intervention, we are not yet in a position to recommend a concentration of interventions downstream in the food chain.The reporting of the cost-effectiveness analysis satisfies project milestone 06/02.

6.3 Estimate of the public health cost of salmonellosis due to STMMethods: Cases were those with a specimen positively identified as STM by the laboratory between July and December 2008. A letter describing the study and invitation to participate in the study was sent to all cases by the HPA at Colindale. Those who returned a recruitment form indicating their willingness to participate were sent a questionnaire and a reminder if no questionnaire had been returned in one month. Ethical approval was granted for this study by the local board. The questionnaire is appended as appendix 3 (available from Defra upon request).

The questionnaire was the source of data about the illness and resources used. Details about the course of the illness were analysed to establish duration and severity of the illness. Use of resources were collated and vectors for the use of NHS resources were costed using the latest NHS reference costs, 2006-7. Direct expenses to cases included costs of transport, medicines (over the counter medicines or prescription costs if appropriate), loss of days of work were estimated and costed using labour force data for 2006-7.

Results The public health costs of a case of Salmonella Typhimurium (STM) were estimated separately for the families of the cases, and to the NHS. Of the cases included in the study 59% were females and 62% were over 35 years of age. Most of patients (51%) were in full time work. Most case, 40%, were in households that consisted of two members.The most frequent symptoms were diarrhoea and abdominal pain. The mean number of days with diarrhoea was 8.5 days (0-85 days) and with abdominal pain was 6.5 days (0-36 days). The severity of the disease had a significant impact on the ability of the patients in carrying on daily activities: 94% of ST patients had spent on average 13.5 days from the beginning of the illness until they were able to carry on normal daily activities. 38% were away from paid work for 4.9 days and 31% were away from planned leisure and/or social activities for 4.2 days. About 69% of ST cases needed someone else to look after them whilst they were ill, so the illness affected not only the patient, but also another person: 23% of these persons took about 1 day off work.

The costs reflected the severity of the disease. The mean direct family cost for ST was £54.91. The indirect cost associated to the lost of income by the case or a career was £182.35. Estimated average NHS cost per patient, was £611.44, where more than half of these costs were with hospitalization (£343.75) of which £118.30 was for time in intensive care. The total cost the patient’s GP surgery visits was £3425.03. This contributed (23%) to total NHS costs (£14,721.56). GP costs were estimated at 2007 reference prices.These results will be combined with the cost-effectiveness data and the MRA to provide a societal cost-benefit analysis, which will be submitted to a peer-reviewed journal and will also inform the economic analysis commissioned by DG SANCO (EU).

6.4 Conduct a full farm to fork stakeholder analysisStakeholder interviews were conducted by telephone or in person (S06/06) with 7 farmers, 7 abattoir managers and 2 veterinarians. This was a smaller group of stakeholders than had been planned, which reflected the experience that it was impractical to conduct lengthy telephone interviews with semi-structured and open questions. Instead, individual face to face interviews were conducted and these were more expensive. There was general support for the implementation of measures to control Salmonella in pigs and pork, reflecting the perceived need to attain good food safety standards in order to generate consumer confidence in food safety and quality. The responsibility for prevention of human infection was perceived to be a consumer responsibility. Stakeholders thought that Salmonella control measures should be adapted and related to individual farm conditions. The costs of implementation and maintaining control impeded implementation as the benefits were perceived to be indirect and unrelated to animal or human health. The voluntary nature of ZAP/ ZNCP was viewed favourably; mandatory control measures were not. In

SID 5 (Rev. 3/06) of 30

conclusion, farmers considered that reduction of Salmonella would be an achievable goal, but elimination was considered to be impossible. A report on the stakeholder analysis was circulated and a scientific paper has been submitted to the Pig Journal for publication, in satisfaction of project milestone 06/03.

INTEGRATION – OBJECTIVE 7

A single workshop was anticipated to mark the initiation of the integration work package (project milestone 07/01). In practice, this was substituted with a series of meetings with the different collaborating groups.

7.1 Extend MRA to include spatial elements derived from Objective 5The odds ratio’s (ROR) for each region of Great Britain and Northern Ireland (GB & NI) estimated from the spatial analysis in objective 5 are integrated with the farm transmission module of the MRA. From GQA (2007), QMS (2007) and ABP (2007) we obtain the number of assured pig farms in each region and use this to estimate the percentage of farms in each region (POR). Regions with similar prevalences have been grouped, in accordance with areas that have similar farm management (finisher flooring systems, herd size etc) and are geographically close. The transmission parameter, for region l, w

R(l), is then defined as

where w is the transmission rate from the baseline model (estimated from objective 4). The sum in the equation is to normalise so that the average transmission over all regions is equal to w. At the beginning of each iteration of the farm model, we determine which region the farm is in according to the probabilities, POR, and use the corresponding transmission parameter.

Figure 4 shows the average prevalence of excretor pigs, by region. We can see, as may be expected, that the average on-farm prevalence varies between regions. Those regions with a high odds ratio have a higher average prevalence than those with a low odds ratio.

Figure 4: Average on farm prevalence of excretor pigs by region

7.2 Use results from objectives 1, 4 and 5 to review and revise parameter estimation in the MRAA microbial risk assessment (MRA) has been developed as part of objective 3. The results from other objectives (i.e. 1, 5 and 6) of the current project have been integrated into the MRA. Data from objective 1, supplemented with literature, has been used to investigate on farm scenarios, such as improved rodent control, organic acids and vaccination. The spatial analysis from objective 5 has been incorporated into the farm module to allow for differentiation in transmission of Salmonella between different geographical areas of the UK. The MRA is instrumental in the socio-economic analysis of objective 6, where the resulting effectiveness measure for each intervention is combined with costs for each intervention and associated human illness. In doing so an overall relative cost-benefit of each intervention is derived. Lastly the decision support tool of objective 8 also relies on output of the farm module of the MRA and various scenarios.

7.3 Integrate results from socio-economic studies (Objective 6) into MRA to allow concurrent modelling of the physical impact of possible interventions and the economic costs/benefits

SID 5 (Rev. 3/06) of 30

The human illness costing from objective 6 estimated that the cost per human case of Salmonella Typhimurium specifically was £848.70. We assume that only 1 in 4 of the cases predicted by the MRA are actually reported. If a case is not reported, then we assume that the person has not been to the hospital and, therefore, the cost of an unreported case is considerably less than that of a reported case. Therefore, for unreported cases, we assume that the cost per case is only the average family direct cost (£54.91).

The results of the integration of the MRA and the cost of human illness are shown in table 7. From this we predict that the annual cost of human salmonellosis attributable to pig meat is £234,276. However, it is important to emphasise that this prediction is derived entirely from our MRA and the true PAF for human salmonellosis is not known. Therefore, the true costs could plausibly be greater or less than this prediction. Further data on PAF would greatly improve the accuracy of our predictions.

Table 7: Predicted annual cost of human Salmonella infection and disease attributable to pig meat

Cost per case

Average no. of cases per

year for chops


year for bacon


year for sausages

Annual cost

Reported cases £848.70 24.1 0.4 125.6 £127,390Unreported

cases£237.26 72.4 1.3 376.8 £106,886

Total - 96.5 1.7 502.4 £234,276

7.4 Use the model in scenario analysisThe effectiveness of various farm and abattoir interventions are investigated by using data from previous studies (e.g. OZO613) and published research to amend the appropriate parameters in the MRA. The resulting change in number of human illnesses between the baseline model and the intervention is calculated. This is combined with an industry cost for each intervention to derive a cost-effectiveness ratio (Table 88).

Table 8: Percentage reduction in human illnesses, compared to the baseline model, and cost-effectiveness ratios of interventions

Intervention Percentage reduction, P(i)

Industry cost in terms of pigs, C(i)

Cost-effectiveness ratio, E(i)

Bunging 34.43% £1,755,000 0.1962Vaccination and

bunging78.17%

£4,706,5310.1661

Organic acids 59.82% £4,450,000 0.1344Vaccination 36.53% £2,951,531 0.1238

Rodent control 69.66% £22,351,412 0.0312Rodent control 1/3

less effective53.57%

£17,921,0880.0299

Washing and drying 11.81% £5,202,000 0.0227Project milestone 07/02 required a draft report on the integration work package to be submitted to Defra in January 2009. However, this was replaced by the inclusion of this information in the Final Report.

ReferencesABP (2007) Assured British pigs assurance scheme.GQA (2007) Genesis quality assurance pig schemeQMS (2007) Quality Meat Scotland

DECISION SUPPORT TOOL – OBJECTIVE 8

8.1 Development of a prototype decision support toolThere was a project objective to develop a decision support tool (DST) that would enable farmers and veterinarians to make informed decisions in controlling the prevalence of Salmonella Typhimurium within slaughter age pigs, on the farm, prior to transport to slaughter. It was envisaged that a web-based interface would have been designed that would allow the user to input values of specific characteristics relating to their holding that were relevant to Salmonella prevalence (e.g. herd size, type of flooring used in pens). The interface would be simple and easy to follow, and would act as the front-end of a database containing extracts from the on-farm transmission model (see Objective 3 and Hill et al., 2003). However, the DST was designed to be relevant for the ZAP scheme, which was replaced by the ZNCP in 2008. It was determined that re-designing the DST for a completely new situation, whilst feasible, would demand a significant increase in finance. Since the ZNCP will most probably be replaced or at least undergo important revision once the EU targets are set, which could once

SID 5 (Rev. 3/06) of 30

more render the DST inappropriate, it was decided instead to cease any further development. Nonetheless, the technical insights gained are considered to be valuable and will inform any future efforts to create such a tool.

MethodologyThere are many different interacting variables that influence the transmission of Salmonella on a farm. Discussion with project team experts yielded a list of farm specific characteristics and interventions that it was thought would be important to include. However, there was a lack of data in respect of some of these, which would require to be substituted by expert opinion or extensive additional field studies. Relating interventions or farm factors to particular parameters such as transmission rates would simplify the process, but would require extensive expert opinion and work in order to identify any and all correlations between parameters.

The front-end of the DST was developed in Microsoft Access. It was designed to be accessible to users with little modelling and scientific knowledge. For each characteristic the user is required to choose from a selection of options that most represents their holding (e.g. slatted or solid flooring). The results from the on-farm transmission model were stored in a database. The DST selects the outputs from the database that relate to the users input and returns them to the user. In the prototype model the outputs are the mean and 95 th percentiles of Salmonella prevalence for each combination of characteristics.

To allow the DST to move forward we need to further establish requirements, data needs and how to incorporate suggested risk factors in the transmissions model and the DST. The requirements of the DST may need to be refocused in such a way as to align with data restrictions and the capabilities of available transmission models. The DST would be extended to take into account the prevalence on the users holding.

DiscussionWe have developed a DST framework which aims to allow producers to input information about their holding and get an estimate of Salmonella prevalence on similar holdings. If known, this can be compared with the prevalence on their holding. The DST is easy to use and has the benefit of allowing the user to change inputs to see how it would affect Salmonella prevalence. Although development of the DST is currently suspended, the concept behind the idea remains fundamentally valid. To this end, as part of a redefined brief, it would be possible for farmers to see the impact, on Salmonella prevalence, of changing one or more elements of their Salmonella control package. The database would display the potential benefits of each change via prevalence and could, with a simple costing study, display cost-effectiveness information for every proposed change.The work completed under this objective and reported here replaced project milestones 08/01 and 08/02 as agreed with Defra.

COMMUNICATION – OBJECTIVE 9

Review pig herd health programmesThrough another collaborative project with University of Newcastle and SAC (OD0215 – “Pig Health”), we obtained details of existing herd health programmes (secondary milestone S09/01) from pig producers. These were reviewed in order to identify their common features, to describe the data that they provided and the utility of this data considered for individual farmers and their veterinary advisers.The eight reviewed plans have much in common as they mostly require answers in open text form, although one collects only answers as numerical valuations. Two other plans use some scoring but in general the information is collected as detailed text answers. Their formats reflect their origins, in that the Farm Assurance and Certification Bodies concentrate on meeting certain standards and biosecurity compliance on the part of the farmers, with less interest in animal health or breeding programmes. Those plans originating from Veterinarians and Consultants emphasise information pertaining to disease, preventive medicine and treatment, but these also collected data on a wide variety of farm management practices, including detailing the training and experience of stockmen. The two remaining plans, drawn up by a manufacturer of disinfectants and a private company, vary in content with the private company plan being similar in design to those created by the vets. However, the disinfectant manufacturer scheme has been designed with a very different purpose in mind, and it collects detailed and comprehensive information on biosecurity and farm cleanliness procedures. In summary, the plans, especially from the vets, contain detailed information which could be useful for analysis; however, the data are collected in a non-standardised manner and the open manner of the questions would make the information difficult to analyse.

9.1 Evaluate opportunities to enhance salmonella control through existing programmesMany farmers invest in herd health visits which are often conducted by expert pig veterinarians. Many of these have gained additional qualifications, for example Royal College of Veterinary Surgeons Diploma or Certificate in

SID 5 (Rev. 3/06) of 30

Pig Medicine. Some farmers will use these experts as consultants, in addition e.g. to a local veterinary practice who provides a “fire brigade” clinical service. Such visits may be carried out at intervals varying from monthly to annually. As described above, Salmonella control is unlikely to yield any direct benefits to the farmer and may be resented as an additional cost.

However, many of the measures that are suggested to control Salmonella may bring about further indirect benefits for the producer. For example, improved biosecurity may be anticipated to reduce the risk of incursion of other infectious agents, such as Brachspira spp, Mycoplasma spp., PRRS virus or PCV-2, all of which may represent important threats to herd health and productivity. Improved rodent control may also reduce risk of other infectious diseases such as leptospirosis as well as bringing about a direct saving in terms of reduced feed losses and damage to buildings and fittings. Farmers consistently list their private veterinarian as their most trusted source of advice. Therefore, if widespread adoption of effective control measures is to be brought about, it will be vital for private veterinarians to become actively engaged in the process and to recognise where measures of multiple utility may be adopted by their clients. Investigations into the association between productivity and prevalence of Salmonella infection should be progressed to provide evidence for the assertion that interventions that are effective for Salmonella also bring about other benefits. Some research is currently underway in this area, through the Defra-funded Pig Health project OD0215. Unfortunately, the current perception that consumers should bear responsibility for food safety (see stakeholder analysis above) may mitigate against these developments.

9.2 Promotion of salmonella controlThe research conducted in this project and previously in OZ0316 has provided evidence for interventions that can have an impact on the prevalence of Salmonella on GB pig farms. We have also identified certain farm characteristics that are associated with a greater risk of infection, which may be valuable for targeting particular sections of the pig industry for action. Our QMRA predicts that all farms must act to some extent if a sufficient reduction in the prevalence of Salmonella in UK pigs to produce an important reduction in human disease is to occur. The evidence which we produced has already been extended to industry via a seminar entitled “Serious about Salmonella” which FSA sponsored in 2008. This was followed up by producing a DVD and booklet plus a series of “roadshows” for farmers. A full revision of the Defra Guidance Notes on the Control of Salmonella in pigs is also planned. Presentations have also been made at scientific and veterinary meetings in UK and at international venues – these are listed in appendix 4.

BPEx has been encouraging producers to undertake interventions with the aim of firstly, showing that control plans can be effective and secondly, of identifying best practice so that it may be promoted via “demonstration farms”. Defra has funded VLA to undertake a study to measure the effectiveness of the BPEx approach in stimulating change in participating farmers (OZ0148). This project builds upon previous work with cattle farmers that yielded a behavioural model, explaining the processes necessary for producers to adopt and sustain behavioural changes to control zoonoses on their farms. The failure of ZAP to provoke any widespread change during the period 2001-2008 emphasises that scientific evidence alone is not sufficient to bring about control and social science approaches, along with economic studies, promise to complement more conventional epidemiological or microbiological research in the future.Participation in the “Serious about Salmonella” workshop satisfied project milestone 09/01, to promote STM control to industry. A presentation was also made to a stakeholder meeting organised by DG SANCO in Brussels in February 2009.

DISCUSSION/CONCLUSIONS

This multi-disciplinary and multi-institutional project has provided a wealth of new knowledge about the epidemiology and control of Salmonella in pigs in GB and the probable impact upon human health. Despite several years of monitoring by ZAP, Salmonella infection remains largely unnoticed by producers as it is not perceived to cause any important economic loss to them. The contraction of the industry and the parlous financial state frequently precludes serious investment and the inevitable deterioration in buildings etc constrain opportunities for producers to tackle Salmonella even if they were minded to do so. However, EC regulation places an inescapable responsibility for safe food production upon farmers and it is anticipated that this will result in a legal obligation to control Salmonella in pigs in the future. UK has an unenviable position as one of the EU Member States with the highest prevalence of Salmonella in slaughter pigs; our home-produced pig meat is more likely to come from infected pigs than that produced e.g. in Denmark, Netherlands or Poland. Whilst at present, there is little public or retail concern about UK pork, an outbreak such as that currently occurring in Denmark (the source of which has not been confirmed but which has seen suspicion fall on the pig sector) could suddenly alter this to the detriment of producers. Therefore, it is essential that evidence-based control strategies are formulated.

Although there is clear evidence that some human cases of salmonellosis do originate from pigs, the evidence is only convincing in respect of the limited number of outbreaks that have been subject to careful epidemiological investigation. Our work in this project has shown that sophisticated microbiological typing methods, e.g. VNTR or

SID 5 (Rev. 3/06) of 30

PGFE, may be valuable in identifying some clusters of Salmonella that are unique to people and pigs but there are also other clusters which have a much wider host range and these methods alone are unlikely to be sufficient to estimate the contribution of pigs to human disease. Uncertainty about the true public health impact of Salmonella in pigs will remain an obstacle to policy formulation and to acceptance of control costs by industry.

VNTR and PGFE were used successfully in this project to show that different strains of Salmonella may arise serially within a single batch of pigs. We have shown that residual infection that was not eliminated by cleaning and disinfection between batches of pigs does lead to re-infection when accommodation is re-stocked. We have also shown that Salmonella introduced with new pigs can spread rapidly through pen mates and to adjacent pens. Finally, we have shown that Salmonella from the farm may be found on the surface of carcasses when the pigs are slaughtered and we have also shown that Salmonella infection encountered for the first time in the abattoir lairage can also result in rapid infection, including of mediastinal lymph nodes and may contaminate carcass surfaces. Thus, benefits from on-farm intervention may be lost during transport and slaughter.

We have used field data to create mathematical transmission models and used these models to estimate the number of pooled samples required to detect infection in various circumstances. This knowledge was utilised in the design of the EU breeding pig baseline survey for Salmonella and will also be applied to the analysis of voluntary additional data from a number of Member States during the 2008 survey.

Our QMRA was extended to include breeding pigs. This module may become increasingly important once the data from the EU baseline survey are available. However, a novel and significant feature of our QMRA was the integration of economic data. It must be emphasised that at present, there is a paucity of valid data for many model parameters and therefore, our results should be interpreted with caution. However, the prediction that abattoir interventions are likely to be more cost-effective than farm interventions in terms of reducing the risk of human disease is extremely important. It is also necessary for decision-makers to recognise that control of Salmonella in pigs may not yield a cost-benefit ration above 1.00 – or in other words, that the societal benefit of a reduced number of human cases may be valued at less than the costs of control that must be borne by industry. This may be expected to provoke debate on the use of private resources to produce public good. Our model offers a systematic approach for consideration of Salmonella control at all levels of the food chain and this advantage will hopefully be useful in considering the optimum disbursal of resources amongst different stakeholders.

In conclusion, there will be increasing pressure to control Salmonella in pigs in the coming 2-5 years. However, the lack of proven interventions on farm and the uncertain net benefits from control will confront decision-makers with a dilemma. It is essential for the maintenance of public and industry confidence that the best scientific evidence is brought to bear, not least to ensure that expectations are effectively managed.

FURTHER RESEARCH IDEAS

Defra is currently funding three projects which involve VLA and that are directly relevant to Salmonella in pigs. OZ0330 is focused upon breeding pigs; OD0215 is more widely concerned with pig health and OZ0148 is considering behavioural change amongst pig farmers.

1. Since control of Salmonella in pigs is predicated on the assumption that these are the origin of an important proportion of human salmonellosis cases, it is essential that an estimate of the Population Attributable Fraction (PAF) for sporadic human disease is made. This parameter is key to all economic analysis and to ensuring that policy requirements are commensurate with risk. Typically, PAF may be estimated by a case-control study.

2. Introduction of Salmonella into the grower herd by infected weaners is a major risk factor for infection at slaughter age. Research should be focused upon firstly, producing the “salmonella-free weaner” and secondly, on development of a rapid real-time test that could be used to assess the Salmonella status of groups of pigs e.g. immediately prior to purchase, at the time of delivery or at the end of a quarantine period. Such knowledge could enable a manager either to strive to maintain freedom for a negative batch of pigs or invest in control e.g. via acidified feed at an early moment to reduce the risk of transmission.

3. Further data on the costs and efficacy of on-farm and abattoir interventions are necessary to enable the QMRA-economic model to be more accurate in predicting the impact of control measures. These data should derive from field and abattoir-based studies.

4. Further studies are needed on the potential for vaccination to control Salmonella in pigs. In particular, it is important to measure the pre-existing burden of infection when vaccination is introduced to determine whether it may overwhelm the pigs vaccine-induced immunity. It is also important to compare the strains encountered in the field with those that are present within the vaccine, to ensure cross-protection.

5. There will be an on-going need for surveillance for Salmonella in pigs and mathematical models could be utilised to optimise approaches, for example with respect to sample type, sample size and diagnostic test. Economic components should be included to ensure surveillance solutions are cost-effective.

SID 5 (Rev. 3/06) of 30

6. Some husbandry systems that are perceived to be welfare friendly, e.g. outdoor rearing, are also associated with an increased risk of infection. Stratification of our QMRA by system would enable an estimation of the overall contribution of these systems to the public health threat to be analysed. This would inform the development of appropriate focused control measures for these higher-risk groups.

ACKNOWLEDGEMENTS

This project has engaged many scientists across all of the institutions that collaborated in the research. We are especially grateful for those who have not been named in scientific articles etc – they were often the key workers, in labs or offices, without whom our work could not have been completed.

We are particularly grateful to all of the producers and abattoir staff who agreed to assist in the study. We recognise that the time and inconvenience that participation caused them did impact upon their daily business.

Whilst this project was funded by Defra, we wish to acknowledge that results from other research funded by FSA also fed into the project.

Finally, we are very grateful to BPEx and ZNCP for their support throughout this long project, especially Derek Armstrong and Veronica Wright.

References to published material9. This section should be used to record links (hypertext links where possible) or references to other

published material generated by, or relating to this project.

SID 5 (Rev. 3/06) of 30

APPENDIX 1 – FSA Project MO1040 Objective 5

APPENDIX 2 – FSA Project MO1040 Objective 5b

APPENDIX 3 – Public Health Costs Questionnaire

APPENDIX 4 – Project OZ0323 Outputs Summary Table

SID 5 (Rev. 3/06) of 30

general enquiries on this form should be made...

Documents