부산대학교 도서관

Despite the success of antiretroviral therapy roll-out, one-million people still die with HIV-infection annually. In high-burden settings, tuberculosis remains the most common proximal cause of hospital admission and death in people living with HIV. In post-mortem series, 90% of fatal HIV-associated tuberculosis is 'disseminated'. This is a form of tuberculosis which has been poorly characterised and, despite the high associated-mortality, never been the subject of interventional trials to define optimal treatment strategies. This thesis contends that the mode of severe HIV-associated tuberculosis is blood stream infection. First it is argued with reference to historical literature that blood stream dissemination is part of the natural history of post-primary tuberculosis infection, and that HIV-associated M. tuberculosis blood stream infection (MTBBSI) can be conceived of as a reversion to, and exaggerated form of this natural history. Using data from a large cohort (n=571) of HIV-infected inpatients with CD4 cell count < 350 cells/mm3 and a new TB diagnosis from Khayelitsha Hospital, South Africa (the KDHTB study), the extent and magnitude of MTBBSI is shown to be a major determinant of clinical phenotype and mortality risk. Systematic, quantitative markers of blood stream dissemination, including TB blood culture, urine-lipoarabinomannan (uLAM), and urine GeneXpert MTB/RIF testing (uXpert), can be combined into a 'disseminated TB score. KDHTB patients have high prevalence of abnormal sodium and fluid balance, metabolic acidosis associated with acute kidney injury, hyperlactataemia, infiltrative liver and splenic pathology, and anaemia. Each of these pathophysiologies in turn correlates to disseminated TB score, and to risk of death, suggesting bacterial burden and MTBBSI are central to the pathophysiology of severe HIV-associated tuberculosis. An individual patient data meta-analysis, with 20 independent data sets comprising over 6000 patients, is used to establish the prevalence of TB blood culture positive disease amongst critically unwell HIV-infected inpatients. This shows that MTBBSI is more common than previous estimates suggest, is a strong independent association with mortality risk, and is also associated with specific increased risk of death if empirical treatment is delayed. The development of tools to identify and measure MTBBSI is described, including Xpert-ultra testing of blood, and the use of a novel dye, DMN-trehalose, to perform direct microscopy on patient blood samples. These techniques are used to provide the first description of the pharmacodynamics of MTBBSI, by serially quantifying blood bacilli load over the first 72-hours of standard TB therapy, in 28 patients with high predicted probability of bacteraemia. In this cohort, risk of mortality is related to several summary measures of MTBBSI dynamics in the first 72-hours of therapy, suggesting this approach can be used to define biomarkers of treatment response. In conclusion, MTBBSI is a highly-specific diagnosis responsible for substantial mortality in hospitalised people living with HIV. Interventions with strengthened bacteriocidal activity, focussed on reducing bacterial burden, are warranted for MTBBSI. Tools developed in this thesis, including potential pharmacodynamic biomarkers, should facilitate such trials.