This thesis focuses on formulating mathematical models describing antiretroviral use, and in particular, on their potential effect in resource-poor settings. Over recent years, many steps have been taken towards rolling out antiretroviral therapy (ART) access on a large scale throughout sub-Saharan Africa, the progress of which is evaluated. The models aim to describe the impact of antiretroviral use in a more realistic way than has been achieved previously by incorporating factors that are important in ART use, such as stages of HIV infection. The thorough literature review for parameterising the ART models has led to a systematic review and meta-analysis of transmission probabilities for sexual HIV transmission, including heterosexual and homosexual per act and per partnership transmission probabilities. The population-level effect of ART is investigated, showing that even with high coverage levels and timely treatment initiation, ART cannot be used as a control measure in the absence of prevention initiatives. Key issues regarding treatment delivery in ART programmes are investigated, examining the impact of treatment initiation by infection stage (at AIDS or before) and by availability of laboratory facilities. This is investigated within the context of a limited capacity programme where demand for treatment exceeds supply and leads to a discussion of the ethical stances that various possible treatment strategies would implicitly represent. An extended model incorporating use of a second line regimen is used to illustrate the epidemiological consequences of providing further, more expensive treatment to those failing the original regimen, which again provokes debate as to the right strategy for distribution of limited resources.