This thesis investigates the clinical impact of a point-of-care diagnostic strategy for pulmonary tuberculosis (TB) in a setting at the heart of the TB and human immunodeficiency virus (HIV) epidemics in rural KwaZulu-Natal, South Africa. Although the identification and prompt treatment of active pulmonary TB disease remains the cornerstone of global TB control strategies, weak diagnostic systems contribute to substantial delays and default during the diagnostic process. As new diagnostic technologies are developed, evidence is needed around how best to deliver them within health systems in order to maximize their impact. The impact of positioning of a molecular diagnostic system (Xpert MTB/RIF) was investigated in a cluster randomised trial. Clusters (two-week time periods) were randomised to one of two strategies: centralised laboratory Xpert MTB/RIF testing or point-of-care Xpert MTB/RIF at the clinic. The trial enrolled 1297 adults with symptoms of pulmonary TB who were HIV infected and/or at high risk of drugresistant TB. There was some evidence that point-of-care placement shortened the time to initiation of treatment but there was no difference in the overall proportion of culture-positive pulmonary TB cases initiated on appropriate anti-TB treatment within 30 days. Overall mortality was lower than anticipated and, although it was higher with the point-of-care strategy, this effect was not maintained after adjusting for the presence of TB disease and CD4+ T-cell count. Further analysis suggested that the point-of-care strategy increased the proportion of valid Xpert results from the initial sputum specimen, increased the proportion of individuals receiving test results and allowed same-day treatment initiation for half of all culture-positive cases that tested positive with Xpert. The diagnostic performance of the Xpert MTB/RIF system was comparable under both strategies. However, delays in initiation of treatment for drug-resistant TB cases and for Xpertnegative/ culture-positive cases occurred similarly with both strategies, reducing the potential to detect a real impact on outcomes. Although not a primary focus of the study, the results highlighted deficiencies in the performance of sputum culture, which raise questions about its place as the gold standard diagnostic test. The development of simple, rapid diagnostics suitable for point-of-care use remains important for TB control in high burden settings. The findings will improve understanding of the key requirements for successful diagnostic strategies and the lessons learnt will help to inform future diagnostic clinical trials. Further research is needed to evaluate how different diagnostic strategies might impact on TB transmission in health care facilities and more broadly in the community.