Observational data are increasingly used in various therapeutic areas to evaluate the use, effectiveness and side effects of new treatments. Whilst randomized clinical trials remain the gold standard for evaluating the efficacy of new agents, they have a number of limitations for HIV, including the limited number of combinations that are compared and the costs of long-term follow-up. Observational data from seroconverter and clinical cohorts have been used to compare the short- and longer-term effectiveness of different therapy combinations and to evaluate the longer-term risks associated with antiretroviral therapy. Furthermore, they provide the opportunity to evaluate the relative success of more complex patterns of therapy, such as sequencing of different regimens over time. However, because of the nature of these data, a number of potential biases may arise which can influence interpretation.