The global burden of tuberculosis (TB) is staggering. In 2017, 10 million people developed TB and 1.6 million people died from TB worldwide. Successful treatment of drug-susceptible TB, defined as the combination of clinical cure or treatment completion, requires at least six months of therapy. Despite widespread availability of such treatment, the global treatment success rate is sub-optimal, estimated at only 82% in 2016. TB/HIV co- infection further complicates treatment of both conditions due to drug-drug interactions, high pill burden, and diminished immune function. However, the degree to which HIV-related disease characteristics, such as CD4 cell count, HIV-1 RNA viral load, and timing of antiretroviral therapy initiation, impact TB treatment outcomes is unclear. To optimize successful treatment outcomes, healthcare providers need simple and effective tools at the start of treatment to identify TB patients at the greatest risk of poor outcomes, and whose prognosis could be improved through tailored care management and treatment monitoring. To address these knowledge gaps, this proposal aims to develop a prediction model for TB treatment outcomes with a focus on characterizing HIV severity. I will also incorporate pharmacogenomic data on acetylation status, informed by known genetic variants of NAT2 that are involved in the metabolism of isoniazid. Isoniazid is one of two TB drugs taken for the entire 6 months of standard treatment, and its efficacy is fundamental to TB cure. The plan to quantify the added value of HIV severity and acetylation status in the prediction model will be useful to understand whether collection of such data improves TB outcome prediction and is worthwhile to collect in clinical settings. This study will leverage existing data from the Regional Prospective Observational Research for Tuberculosis (RePORT)-Brazil project, an observational cohort co-funded by the NIH and Brazilian Ministry of Health that has enrolled 940 culture-confirmed, pulmonary TB cases in Brazil. The proposed research will address a fundamental question about what combination of clinical, epidemiologic, and pharmacogenomic factors are best able to predict TB treatment outcomes. It will support future patient-centered approaches in TB therapy. Additionally, the innovative training plan will foster my academic development as I further my predoctoral training in epidemiology at Vanderbilt University. The combination of strong and dynamic mentorship, relevant TB-specific knowledge, and focused methodologic training in prediction modeling and pharmacogenomic data analysis will provide a solid foundation to build a successful career as an independent investigator with expertise in TB epidemiology.
Previous prediction models for TB treatment outcomes have considered HIV status as a binary covariate in populations of TB patients living with and without HIV but none, to my knowledge, have characterized the influence of HIV severity, such as CD4 cell count and HIV-1 RNA viral load, or timing of antiretroviral therapy initiation. Additionally, more recent research suggests that pharmacogenomics (primarily isoniazid acetylation status) play an important role in TB treatment outcomes, but such data have not been incorporated into prediction models alongside clinical and demographic predictors. To address these knowledge gaps, this study proposes using rigorous statistical methods to develop a model for predicting TB treatment outcomes with a focus on characterizing HIV severity and assessing the added value of isoniazid pharmacogenomic data to the prediction model.
