Model-based tools to improve treatment of tuberculosis patients
Tuberculosis (TB) infects 10 million people each year and kills more than any other infectious disease. All current approaches to TB treatment are based on a one-size-fits-all approach, which leads to undertreatment of patients with severe forms of disease and entails unnecessarily long treatment with potential toxicities for many patients in whom the disease is less severe. Shorter, efficacious, and better-tolerated oral regimens for TB are needed. Unfortunately, all recent Phase 3 clinical trials aimed to shorten treatment duration from 6 months to 4 months for drug susceptible TB failed. The aims of this dissertation were to quantitively characterize treatment response in TB patients and develop model-based tools that provide informed recommendations on optimal treatment regimens and strategies that: i.) maximize durable cure in all patients, ii.) maximize success of late stage regimen development, and iii.) minimize safety concerns associated with a highly potent, but toxic, high-dose linezolid-containing regimen.
In a patient-level pooled analysis of all recent Phase 3 clinical trials evaluating shorter treatments for drug susceptible TB , survival analysis identified risk factors of treatment outcomes. Based on these risk factors, a risk stratification algorithm and clinical simulation tool were developed to provide more individualized predictions of optimal treatment regimens to achieve high cure rates in TB patients.
TB regimen development is plagued with many challenges, the most serious being the inability to identify optimal regimens early and efficiently. To facilitate decisions on novel TB regimens that move forward through the development process, an integrated model was developed to describe the translational link between Phase 2 intermediate biomarkers, treatment characteristics, and patient risk factors to Phase 3 clinical outcomes. We provide clinical trial simulation tools to design innovative clinical trial designs that permit evidence-based decisions on moving the best regimens forward in late stage clinical development .
The TB regimen development process is also challenged by the lack of reliable, quantitative, non-culture-based biomarkers that inform individual level and trial level treatment response. We showed that leveraging longitudinal sputum culture results and drug exposure and applying advanced nonlinear mixed effect modeling with machine learning approaches offer insights into the response dynamics following anti-TB treatment. We identified candidate proteomic signatures that can potentially predict treatment response.
Lastly, using modeling and simulation approaches, we quantified the pharmacokinetic-toxicodynamic relationship of a high-dose linezolid-containing regimen for extensively drug resistant TB. We provide practical data-driven recommendations about linezolid dosing adjustments to optimize therapeutic effects and minimize adverse events.
The quantitative, model-based tools developed in this dissertation contributes to providing evidence-based recommendations on optimal treatment strategies for current and novel TB regimens.