This thesis is concerned with the molecular epidemiology of Campylobacter jejuni and Salmonella in New Zealand and the development of source attribution tools for these pathogens. Although campylobacteriosis is the leading enteric zoonosis worldwide, the pathogen's complex epidemiology and di culties with existing typing schemes, have posed challenges for the control of this disease. The rst study of this thesis gives an overview of existing approaches to microbial risk assessment and source attribution, with particular respect to campylobacteriosis, and describes their advantages and shortcomings. Further, the chapter discusses phenoand genotyping techniques for Campylobacter spp. and the value of including microbial typing data in risk assessments. In the second study, data from a sentinel surveillance site in the Manawatu region was used to investigate the molecular epidemiology of human campylobacteriosis cases. This analysis revealed the presence of a dominant C. jejuni clone, namely sequence type (ST) 474, which accounted for 30.7 % of human cases in the study and identi ed risk factors for infection with ruminant and poultry associated STs. The third study investigated the link between C. jejuni in human cases and samples taken from poultry. By applying epidemiological and population genetic techniques this part of the thesis provided further evidence that poultry is a major contributor to human infection. In the fourth study an existing Bayesian source attribution model was modi ed and consecutively applied to New Zealand's major foodborne zoonoses: campylobacteriosis and salmonellosis. The majority (80 %) of human campylobacteriosis cases attributable to C. jejuni were estimated to have been acquired from poultry sources, whereas wildlife source were estimated to contribute only a minor proportion of cases. In the fth study the Salmonella dataset was descriptively analysed and a large proportion of human cases was found to be caused by `exotic' Salmonella types. In the nal study of this thesis four di erent genetic and epidemiological source attribution methodologies were applied to the same dataset in a comparative modelling framework. iv The studies in this thesis show that epidemiological studies combined with molecular tools and modeling can provide valuable risk-based tools to inform the surveillance and control of zoonotic pathogens. Methods from these studies may be readily applied to the control of other (food borne) zoonoses and provide new opportunities for epidemiological investigations and source attribution modelling of major pathogens.

Johne's disease (JD), caused by Mycobacterium avium subspecies paratuberculosis (MAP) is a chronic, debilitating enteritis of cattle, other domestic livestock and some wildlife species. JD was first identified in the late 1800s and today it is a worldwide problem in dairy cattle. Heavily infected cows have reduced milk production, a higher risk of removal from the herd and low slaughter value. Several countries have implemented national level control strategies. In New Zealand, JD was first reported in 1912 and today the prevalence of infected dairy herds is thought to be high. To improve our understanding of the epidemiology of JD and to evaluate the feasibility of a national control strategy, four studies were conducted. The first study was a questionnaire based case-control study to identify associations between management practices and the occurrence of clinical JD on farms from four regions of New Zealand. The second study was on the effect of sub-clinical JD on milk production and the risk of removal from the herd in four dairy herds over four milking seasons. The effect of misclassification of disease status on productivity estimates was also studied. In the third study diagnostic test result data from the productivity study was combined with a novel Bayesian regression model to estimate performance of the ELISA and faecal culture tests as a function of covariates and utilising repeated tests on individual cows. Finally, results from these three studies were used to adapt an existing JD simulation model, 'JohneSSim', to represent the epidemiological behaviour of JD in New Zealand dairy herds. Control strategies for the disease were simulated and evaluated based on their cost effectiveness. Of the 427 farmers responding to the questionnaire, 47% had suspected clinical cases of JD in their herd in the preceding 5 years. Only 13% of suspected infected herds had an average incidence of greater than 0.5 cases per 100 cow years at risk. The disease was not considered a serious problem by 20% of herd managers who reported the presence of disease in the preceding 5 years. The presence of Jersey cows in the herd and the purchase of bulls had strong positive associations with the presence of clinical JD. Grazing calves in the hospital paddock, larger herds, the purchase of heifers, and the use of induction were also positively associated with JD. In the productivity study the herd-level prevalence of JD by ELISA and/or faecal culture ranged from 4.5% (95% CI 2.6-6.9) to 14.2% (95% CI 9.2-20.6). Daily milksolids production by JD positive cows was 0.8% (95% CI -6.1%-4.5%) less than that of JD negative cows. However in herd D, JD positive cows produced 15.5%, (95% CI 6.75%-24.2%) milksolids less than JD negative herd mates daily. This equates to a loss of 53kg of milksolids/305 day lactation, or NZD 265/lactation, given a price of NZD 5/kg of milksolids. In herd D only, the annual hazard ratio of removal for JD positive cows was significantly increased. It was 4.7 times and 1.4 times higher in cows older than 5 years and younger than 5 years. The results were insensitive to misclassification. Analysis of the diagnostic test data demonstrated the strengths of our Bayesian regression model. While overall estimates of sensitivity and specificity by this method were comparable to estimates by existing methods, it showed a broad trend of increasing sensitivity in higher parity groups and higher sensitivity in early, relative to late, lactation. It also showed that estimates of prevalence may in fact decline with repeated, relative to single, testing. Our novel approach demonstrated trends that could not be shown by existing methods, but could be improved by application to a larger data set. Simulation showed that control strategies for JD based on either test-and-cull, vaccination, breeding for genetic resistance, or removal of offspring from clinically affected cows, were not cost effective for the average infected herd. Improvement of the hygiene associated with calf management provided the greatest reduction in the within-herd prevalence of JD. While JD is present in a high proportion of New Zealand dairy herds, the incidence of clinical cases is usually low, and most farmers consider it to be of little importance. However, JD causes significant losses in productivity in some herds. The disease would probably be best controlled on a herd-by-herd basis, given the limited success of national-scale control programs for JD in other countries. The education of dairy farmers regarding risky management practices, and the offer of a risk assessment to farmers wishing to control the disease, would provide a combination of wide reaching and targeted approaches, of low cost, for JD control. It seems likely that JD will persist in some capacity in the years ahead, but will remain of minor concern next to major animal health issues, such as infertility and mastitis. Clarification of the effect of genetic strain on the virulence of MAP may help explain differences in the effect of the disease between herds. This knowledge could then be used to further improve the efficiency of JD control.

The eco-epidemiology of bovine tuberculosis (Tb) in wild deer (mainly red deer Cervus elaphus) in New Zealand was investigated. Bovine Tb is caused by Mycobacterium bovis. Specific aims were to clarify the likely routes of infection in deer, and to determine the status of deer as hosts of Tb, the likely rates and routes of inter- and intra-species transmission between deer and other wildlife hosts, the role of deer in spreading Tb, and the likely utility of deer as sentinels of Tb presence in wildlife. As the possum (Trichosurus vulpecula) is the main wildlife host of Tb, the research also included some investigation of transmission routes in possums. Patterns of infection were measured in 994 deer killed between 1993 and 2003. Tb prevalence varied between areas (range 8–36%). Few deer had generalised infection, with 21–68% of infected deer having no visible lesions, depending on the area. The retropharyngeal lymph nodes and oropharyngeal tonsils were commonly infected. No dependent fawns less than 0.75 years old were infected, indicating intra-species transmission is rare in wild deer. Where possums were not controlled, the net (cumulative) force of infection in young (1–4 y) deer was 0.10–0.24 per year in males and 0.09–0.12 per year in females, but much lower in older deer (less than 0.05 per year). Possum control reduced the net force of infection quickly, and eventually to zero. However, Tb persisted in possum-controlled areas through immigration of infected deer and, for almost a decade, through the survival of resident deer infected before possum control. Tb was lost from infected deer at an exponential rate of 0.13 per year, mostly as a result of deer recovering from infection rather than dying from it. Wild deer do die of Tb, but there was no discernible effect on age structure. The occurrence of infection in deer was not linked to the local deer or possum density at their kill sites (i.e. in their home range), but the area-wide prevalence of Tb in deer was closely correlated with Tb levels in possums, which were in turn correlated with area-wide measures of possum density. For wild deer in New Zealand, Tb is a persistent but usually inconsequential disease of the lymphatic system. It is acquired mainly by young independent deer, usually orally via the tonsils, and probably as a result of licking infected possums. Many species fed on deer carrion, including possums. Most possums encountering carrion did not feed on it, but a few fed for long periods. Other scavengers such ferrets (Mustela furo), hawks (Circus approximans), and weka (a hen-sized flightless native bird; Gallirallus australis) fed in a way that probably increased the infectivity of carrion to possums. Commercial deer hunting may have facilitated the historical establishment of Tb in possums. Scavenging (including cannibalism) and interactions with dead and dying possums are identified for the first time as potentially important routes for transmission of Tb to possums, and I develop new hypotheses involving peri- and post-mortem transmission in possums that explain many of the epidemiological patterns that are characteristic of the disease in possum. In continuous native forest, deer home range size averaged 250 hectares for six young females, and over twice that for two males. Over 90% of infected deer are likely to die within 2 km (females) or 6 km (males) of where they acquired Tb, but deer could occasionally carry Tb up to 30 km. Deer will be useful as sentinels, but only where other sentinels are rare, because the force of infection for a deer with a single infected possum in its home range is only 0.004 per year, compared to greater than 0.2 per year for deliberately released pigs. Deer are occasionally capable of initiating new cycles of infection in wildlife, but deer control is not essential to eradicate Tb from wildlife.