It is known that antiretroviral drugs can induce immunologic improvement in patients with acquired immunodeficiency syndrome (AIDS) and other manifestations of HIV infection. However, the improvements so attained are often partial and transient. This may result from a number of factors, including incomplete inhibition of human immunodeficiency virus (HIV) replication by available agents, the development of viral drug resistance, the effect of cytokines, or thymic damage. An understanding of this problem may be important in further development of AIDS therapies. It will also be important to learn how to best assess the response to various therapies. In this regard, the CD4 count is evolving as a mortality risk indicator in AIDS and as such may find utility in assessing new therapeutic approaches. We have observed that in a cohort of gay men receiving antiretroviral therapy in a research environment, nearly all deaths occurred in individuals with fewer than 50 CD4 cells/mm3. However, the relationship between the CD4 count and the hazard of dying may be influenced by a number of factors (e.g., active intravenous drug use, extreme poverty, etc.), and further studies are needed to define the relationship between CD4 and clinical endpoints under a variety of conditions.