Investigating the Epidemiology of bovine Tuberculosis in the European Badger

Abstract

Global health is becoming increasingly reliant on our understanding and management of wildlife disease. An estimated 60% of emerging infectious diseases in humans are zoonotic and with human-wildlife interactions set to increase as populations rise and we expand further into wild habitats there is pressure to seek modelling frameworks that enable a deeper understanding of natural systems. Survival and mortality are fundamental parameters of interest when investigating the impact of disease with far reaching implications for species conservation, management and control. Survival analysis has traditionally been dominated by non- and semi-parametric methods but these can sometimes miss subtle yet important dynamics. Survival and mortality trajectory analysis can alleviate some of these problems by fitting fully parametric functions that describe lifespan patterns of mortality and survival. In the first part of this thesis we investigate the use of survival and mortality trajectories in epidemiology and uncover novel patterns of age-, sex- and infection-specific mortality in a wild population of European badgers (Meles meles) naturally infected with Mycobacterium bovis, the causative agent of bovine tuberculosis (bTB). Limitations of dedicated software packages to conduct such analyses led us to investigate alternative methods to build models from first principles and we found the NIMBLE package to offer an attractive blend of flexibility and speed. We create a novel parameterisation of the Siler model to enable more flexible model specification but encounter the common problem of competing models having comparable fits to the data. Multi-model inference approaches can alleviate some of these issues but require efficient methods to carry out model comparisons; we present an approach based on the estimation of the marginal likelihood through importance sampling and demonstrate its application through a series of simulation- and case-studies. The approach works well for both census and capture-mark-recapture (CMR) data, both of which are common within ecological research, but we uncover challenges in recording and modelling early life mortality dynamics that occur as a result of the CMR sampling process. The final part of the thesis looks at another alternative approach for model comparison that doesn’t require direct estimation of the marginal likelihood, Reversible Jump Markov Chain Monte Carlo (RJMCMC), which is particularly efficient when models to be compared are nested and the problem can reduce to one of variable selection. In the final chapter we carry out an investigation of age-, sex-, infection- and inbreeding-specific variation in survival and mortality in a wild population of European badgers naturally infected with bovine Tuberculosis. Using the methods and knowledge presented through the earlier chapters of this thesis we uncover patterns of mortality consistent with both the mutation accumulation and antagonistic pleiotropy theories of senescence but most interestingly uncover antagonistic pleiotropic effects of inbreeding on age-specific mortality in a wild population for the first time. This thesis provides a number of straightforward approaches to Bayesian survival analysis that are widely applicable to ecological research and can offer greater insight and uncover subtle patterns of survival and mortality that traditional methods can overlook. Our investigation into the epidemiology of bovine Tuberculosis and in particular the effects of inbreeding have far-reaching implications for the control of this disease. This research can also inform future conservation efforts and management strategies as all species are likely to be at increasing risk of inbreeding in an age of dramatic global change, rapid habitat loss and isolation.