The design, conduct, and analysis of clinical trials that evaluate the safety and efficacy of treatment interventions in patients with HIV infection provide many scientific challenges. A recently completed randomized trial of didanosine (ddI) and zalcitabine (ddC), sponsored by the Terry Beirn Community Programs for Clinical Research on AIDS (CPCRA), is an especially valuable resource for illustrating these challenging issues and for providing insights into how they might be properly addressed. Establishing equivalence of treatment effects on clinical efficacy end points is illustrated through the use of the confidence interval approach. The striking changes in treatment efficacy results that occurred during the course of the CPCRA trial provide important insights into how a data and safety monitoring board can reduce the risk of inappropriate early study termination. The trial also provides valuable insights into how treatment effects should be assessed, revealing inconsistencies between effects on the CD4 surrogate end point and effects on primary clinical efficacy end points and showing the incompleteness of the standardly employed definition of AIDS progression. Finally, the results of this ddI/ddC trial are used to examine the role of covariate adjustment.