Incorporating time to response into dose-response models used in quantitative microbial risk assessment

Abstract

While microbial risk assessment has been used for over 25 years, traditional dose-response analysis has predicted only the overall risk of adverse consequences from exposure to a given dose. An important issue for consequence assessment from microbiological exposure is the distribution of cases over time due to the initial exposure. In this dissertation work, the classical exponential and beta-Poisson dose-response models were firstly modified to include dependencies of time postinoculation to quantify the time effect on dose response. By using the maximum likelihood estimation approach, the resulting time-dose-response models were found capable of providing statistically acceptable fits to more than 20 sets of animal survival data administered with graded doses of various pathogens including Bacillus anthracis, Yersinia pestis and Francisella tularensis.The suggested models were further developed to model quantitatively the hypothetical relationship between the host response and instantaneous microorganism number in vivo.A direct verification was conducted using data of monkey inhalation tularemia. The candidate models were fit to survival dose-response data and the estimates of bacterial dynamics for different aerosol sizes were inferred. The predicted growth curve for aerosolized Francisella tularensis was highly consistent with the available bacterial count data.To better prepare for public health consequences following an outbreak event, the suggested time-dose-response models derived from animal data must be verified against human outbreaks. In later part of this thesis, eight sets of survival dose-response data of various animals exposed inhalationally or intranasally to Bacillus anthracis were fit by the proposed models. The parameter estimates were generally consistent across different species. The low-dose time-to-response distributions estimated by most animal models were clearly aligned with the data from the Sverdlovsk outbreak. The models were further applied to quantify the effect of multiple exposures to scrapie agent. The results show that the dispersion of dosing, either temporally or quantitatively, reduces the risk of host response to each single challenge.The outcomes of this work will advance our understanding on dose- and time-dependent probability distributions of disease risks, in vivo pathogenic kinetics after exposure to infectious agents, and the prediction of human epidemic curve based on available experimental data.%%%%Ph.D., Environmental Engineering – Drexel University, 2010