Assessment of Antimicrobial Treatment Strategies in Pig Production Using Mathematical Models

The phenomenon of antimicrobial resistance (AMR) is well established, but its persistent increase and alarming proportions threaten the ability of antimicrobials to treat infections. AMR has become a major issue in veterinary antimicrobial use, specifically in food production animals, due to the potential consequences for human health. Danish pig production accounted for 76% of the total veterinary use of antimicrobials in 2012 with 79% of pig production used in weaning pigs. Escherichia coli (E. coli) are the predominant bacteria in the gastrointestinal flora of humans and animals, and can serve as a reservoir of AMR. Furthermore, around 40% of E. coli isolates from healthy pigs throughout the Denmark over the past five years were found to be resistant to tetracycline and ampicillin. There is, therefore, a need to reduce the levels of resistance in the pig production system using improved treatment strategies. Dosing factors, along with the in vivo epidemiological parameters, govern the relation between resistance and antimicrobial use. Mathematical modeling and simulation techniques have been used over the past two decades to evaluate the effect of these factors on the development of resistance, and are considered to be powerful tools in designing treatment strategies.
The overall aim of the thesis was to develop an in vivo bacterial growth model to predict and assess the effect of dosing factor on resistance growth in order to optimize treatment strategies. Specific aims were to a) estimate pharmacodynamic (PD) parameters of E. coli strains representative of the Danish porcine E. coli strains, when these are exposed to tetracycline and ampicillin, b) characterize the PD effect of combined concentrations of tetracycline and ampicillin, and c) evaluate the treatment strategies that better suppress the growth of resistant strains both under single and multidrug treatments.
Fifty E. coli strains were randomly selected from 160 collected isolates from pigs as a part of DANMAP, and were considered to be representative of the Danish pig population. In vitro growth experiments were performed using BioScreen under exposure of tetracycline and ampicillin, both independently and in combination. PD parameters of strains were estimated for these exposures. Differential equation model use developed for continuous changes in bacterial counts over time in pig intestine both with and without antimicrobial treatment. A total period of 35 days after first day of treatment was simulated in the model. Antimicrobial treatments were introduced in the model based on different combinations of dosing frequency and treatment durations. In addition, the effect of different numbers of competing strains, composition of strains, and increased excretion of strains on growth and levels of resistant strains were assessed.
It was found that the complex relationship between antimicrobial concentration and bacterial growth from in vitro experiments was better explained by PD parameters than single point estimate minimum inhibitory concentration (MIC). These parameters along with MIC should be taken into account when studying the effect of antimicrobials on the bacterial growth. These parameters were used as an input to the in vivo growth model of multiple bacterial strains. For almost all treatments, high resistance levels were found at the end of the treatment, due to the competitive advantage under drug concentration. The speed at which resistance levels fell and returned to the original levels at the end of treatment was then investigated. Short treatment duration (3 days) was found to be beneficial irrespective of the dosing frequency, as long as the dosing frequency remained below a threshold value. Above that threshold, the more frequent the treatment, the more it selected the resistant strains. Besides these dosing factors, the number of competing strains had an effect on the level of resistance. At the end of treatment, resistance levels were the same for different numbers of competing strains, but this level fell more rapidly where fewer strains were competing. Elimination of strains by excretion through feces was found to have large influence on resistance levels. Where more excretion occurred during the treatment, it was more likely that resistance would reach a higher level, with the consequence of a longer return to equilibrium after treatment. For multidrug treatments, sequential treatments were found to better suppress the growth of resistant strains than combination or mixing treatment if introduced in specific order.
Finally, mathematical modeling and simulation provided an excellent opportunity to study and assess the growth dynamics of multiple stains in the intestinal flora of pigs, following both single and multidrug treatments. The model developed in this study is generic and could be used for other drugs as well as in other routes of drug administration. Furthermore, epidemiological parameters were found to have a more profound influence on growth dynamics than dosing regimens. Profiling of the bacteria (using BioScreen or similar to obtain growth characteristics) may be useful in the future before designing treatment regimens. Distribution of bacteria may depend upon the individual pig or herd, and the same treatment could produce different results in different herds or sections of the herd. Varying the treatment regimens for different herds or sections of a herd is a possible step that should be taken after careful field trials.