Epidemiological and ecological approaches to the study of antimicrobial resistance

To date, investigation and policy development relating to antimicrobial resistance (AMR) have focused largely on describing patterns of resistance to individual antimicrobials, often from restricted data sources. What this approach fails to accommodate is the complexity of AMR, such as genetic linkage of resistance determinants, which could potentially lead to inaccurate inference. Novel approaches are therefore required to bring new perspectives on the epidemiology of AMR. These studies take an ecological perspective of AMR, examining several key issues: The relative contribution to AMR from animal and human populations; the potential differences between passive and active surveillance of AMR; the associations of age, antimicrobial treatment and a shared environment with AMR diversity; the relationships between AMR phenotype and genotype; and exploring the additional information provided by DNA sequencing data. Novel ecological and epidemiological approaches were developed to examine long-term passive surveillance data of AMR in Salmonella Typhimurium DT104 (DT104) isolates from concurrently sampled and sympatric human and animal populations in Scotland. By examining the diversity and the phenotypic and temporal relatedness of the resistance profiles, the most likely source of resistance was assessed. The conclusions were that whilst ecologically connected, animals and humans have distinguishable DT104 communities, differing in prevalence, linkage, and diversity. Furthermore, the sympatric animal population is unlikely to be the major source of resistance for human DT104. As robust data are critical to any analysis, the potential differences between data collected by different surveillance types were examined. These systems are not generally designed to detect emerging resistances. The diversity of phenotypic AMR from passive and active surveillance data of poultry Salmonella Heidelberg and swine Salmonella Typhimurium var. 5- isolates derived from animals and foods-of-animal origin were contrasted to assess their suitability for detecting emerging resistance patterns. Results indicated that active and passive surveillance approaches were potentially sampling from distinguishable microbiological communities and are therefore complementary, and that passive surveillance is better at detecting rare profiles. The ecological and epidemiological approach was also applied to a different organism, in a different ecosystem. In order to assess the association of age and antimicrobial exposure with AMR diversity, E. coli from cows and calves on seven dairy farms were examined. There were distinguishable populations of resistance phenotypes on the farms, associated with both age and treatment. Multivariable models were developed to examine simultaneously the association of age, treatment, time, and farm with AMR diversity. These results indicated that there may be particular animal husbandry or farm management practices which influence AMR diversity, and which appear to be different for co-habiting young and adult dairy cattle. The majority of AMR data collected by surveillance systems are phenotypic in nature. However, it is often the underlying genotype that is of interest, which until now could only be achieved with the application of molecular methods. A novel latent class Bayesian model was developed to infer the prevalence of various AMR determinants in a population given a sample of phenotypes, which was applied to animal and human DT104 data. Differences were demonstrated in the estimated prevalences of a number of AMR determinants between the two populations, further supporting the previous observations that the epidemiologies of the organism, the resistance determinants, or both, are distinguishable. To obtain a greater resolution with which to compare AMR in different populations, a second-generation sequencing platform was used to obtain DNA sequencing data from select animal and human DT104 isolates. The objectives were to determine the diversity of the bacteria and of the resistance determinants. Whilst analysis of the resistance determinants is ongoing, preliminary results have suggested that the subtypes of DT104 infecting animals and humans are indeed similar. Overall, these studies comprise application of novel methods and frameworks for the analysis of AMR. The implication of these studies is that greater and more explicit thought is needed regarding the design, collection, analysis, and interpretation of AMR data properly to inform policy