Multiscale Modeling of Tuberculosis Disease and Treatment to Optimize Antibiotic Regimens

Abstract

Tuberculosis (TB) is one of the world’s deadliest infectious diseases. Caused by the pathogen Mycobacterium tuberculosis (Mtb), the standard regimen for treating TB consists of treatment with multiple antibiotics for at least six months. There are a number of complicating factors that contribute to the need for this long treatment duration and increase the risk of treatment failure. Person-to-person variability in antibiotic absorption and metabolism leads to varying levels of antibiotic plasma concentrations, and consequently lower concentrations at the site of infection. The structure of granulomas, lesions forming in lungs in response to Mtb infection, creates heterogeneous antibiotic distributions that limit antibiotic exposure to Mtb. Microenvironments in the granuloma can shift Mtb to phenotypic states that have higher tolerances to antibiotics. We can use computational modeling to represent and predict how each of these factors impacts antibiotic regimen efficacy and granuloma sterilization.

In this thesis, we utilize an agent-based, computational model called GranSim that simulates granuloma formation, function and treatment. We present a method of incorporating sources of heterogeneity and variability in antibiotic pharmacokinetics to simulate treatment. Using GranSim to simulate treatment while accounting for these sources of heterogeneity and variability, we discover that individuals that naturally have low plasma antibiotic concentrations and granulomas with high bacterial burden are at greater risk of failing to sterilize granulomas during antibiotic treatment. Importantly, we find that changes to regimens provide greater improvements in granuloma sterilization times for these individuals. We also present a new pharmacodynamic model that incorporates the synergistic and antagonistic interactions associated with combinations of antibiotics. Using this model, we show that in vivo antibiotic concentrations impact the strength of these interactions, and that accounting for the actual concentrations within granulomas provides greater predictive power to determine the efficacy of a given antibiotic combination.

A goal in improving antibiotic treatment for TB is to find regimens that can shorten the time it takes to sterilize granulomas while minimizing the amount of antibiotic required. With the number of potential combinations of antibiotics and dosages, it is prohibitively expensive to exhaustively simulate all combinations to achieve these goals. We present a method of utilizing a surrogate-assisted optimization framework to search for optimal regimens using GranSim and show that this framework is accurate and efficient. Comparing optimal regimens at the granuloma scale shows that there are alternative regimens using the antibiotic combination of isoniazid, rifampin, ethambutol and pyrazinamide that could improve sterilization times for some granulomas in TB treatment. In virtual clinical trials, these alternative regimens do not outperform the regimen of standard doses but could be acceptable alternatives. Focusing on identifying alternative regimens that can improve treatment for high risk patients could help to significantly decrease the global burden for TB. Overall, this thesis presents a computational tool to evaluate antibiotic regimen efficacy while accounting for the complicating factors in TB treatment and improves our ability to predict new regimens that can improve clinical treatment of TB.