Predicting optimal antibiotic treatment regimens

Resistance to antibiotics is rising, and some infections are not treatable at all anymore. This development threatens the gains in life-expectancy that we have made in the recent decades. One part of the problem is that we struggle with finding new antibiotics, also because developing new antibiotics is very expensive and takes a long time. At the very beginning of antibiotic development, thousands of drug candidates are tested, and in each development step from experiments in test tubes to experiments with animals to several phases of clinical trials the majority of these candidates fail, leaving only 1-2 drugs that can be used in patients. Finding ways to reduce the amount of trial-and-error would help save time and money. We have developed mathematical models that can help reducing trial-and-error in antibiotic development. Together with partners from industry (GlaxoSmithKline) and an international network of university-based researchers (Yale, Harvard, University of Cape Town, US National Institutes of Health, the Leibniz Society, Radboud University, ETH Zurich, Simon-Fraser University), we apply these models to improve therapy of tuberculosis, the bacterial disease that causes most deaths world-wide.We summarized new developments in mathematical modeling in Clarelli et al., 2019, and expanded the toolbox for modeling antibiotic action in Martinecz et al., 2019 and Tran et al., 2020. We have developed easily available mathematical models to estimate the required length on antibiotic treatment (Martinecz & Abel zur Wiesch 2018, Martinecz et al., 2020, Martinecz et al., in revision PLOS Comp Bio), as well as models that can predict bacterial susceptibility to new antibiotics (Clarelli & al., 2020). We also show how such models change predict the emergence of antibiotic resistance (Hemez et al., 2022). In addition, we helped improve mathematical analyses often used in microbiology (Mahmutovic et al., 2020), investigate how EHEC infects patients (Warr et al., 2019) and how antibiotic treatment affects the microflora (Tepekule et al., 2019). We also discussed how to improve drug compliance in patients with tuberculosis (Stagg et al., 2020). Finally, we used our models to help individual TB patients (Koehler et al., 2021)

-

Resistance to antibiotics is rising and new antibiotics are urgently needed. Due to a lack of mechanistic understanding, drug development involves costly trial-and-error approaches to find effective dosing regimens. This makes the development of drugs with low expected revenue, such as antibiotics, difficult. Also, recent clinical trials have shown that dosing of decade-old treatment regimens is suboptimal. We recently developed a new a computational model for the prediction of optimal dosing of antibiotics. This proposal focuses on tuberculosis; the bacterial infection that claims most lives and has the largest disease burden world-wide. Specifically, the aim of this proposal is to contribute to three major goals of improving tuberculosis therapy: i) find optimal dosing strategies (dose levels, dosing frequency), especially for bacteria resistant to first-line drugs; ii) shorten treatment such that it can be more easily adhered to; iii) minimize de novo resistance evolution. In a truly translational approach, we will link this computational model across scales to experimental data, clinical data from 6 clinical trials (among them one currently in press with NEJM) and internal preclinical and clinical data from GlaxoSmithKline. Although this work focuses on tuberculosis, the concepts can be applied to other bacterial diseases and even cancer, as highlighted by our collaboration on cancer therapy with Astra Zeneca.