Statistical Issues Arising in AIDS Clinical Trials
Ellenberg, Susan S., Finkelstein, Dianne M., Schoenfeld, David A., Journal of the American Statistical Association
The evaluation of treatments for the acquired immune deficiency syndrome (AIDS) and human immunodeficiency virus (HIV)-related disorders has become a large-scale effort in a very short period. AIDS as a disease entity was first noted in 1981, when the Centers for Disease Control (CDC) reported rare forms of pneumonia and skin cancer in young gay men in California and New York (CDC 1981 a). The first "definition" of AIDS, developed by the CDC (1982), appeared in 1982. The virus was isolated in 1983 (Barre-Sinoussi et al. 1983), and tests for the disease-documenting antibodies became available the following year. These developments led to a revised CDC definition of AIDS to include a positive antibody test and a more detailed description of the clinical events (primarily opportunistic infections and malignancies) that formed the syndrome.
The identification and isolation of HIV permitted the initiation of drug development efforts. In 1986 funds were first awarded by the National Institute of Allergy and Infectious Diseases (NIAID) for a national cooperative clinical trials program modeled loosely on the oncology cooperative group model developed at the National Cancer Institute (NCI). The efforts to find a treatment produced the first effective drug, zidovudine (ZDV or AZT), in a surprisingly brief period; The results of a highly positive randomized placebo-controlled trial of AZT were published early in 1987 (Fischl et al. 1987), and the drug was approved for marketing by the Food and Drug Administration (FDA) shortly thereafter.
AIDS is different from other life-threatening diseases for which treatments are routinely studied in clinical trials. First, it is an epidemic. AIDS is an infectious disease for which a substantial proportion of the population is at risk. This circumstance provokes a different mind-set than that associated with other frequently fatal diseases, such as cancer, that affect far greater numbers of individuals. This infectious disease, however, is also a chronic illness. The sense of urgency engendered by the concept of an epidemic is therefore directly confronted with all the complications and delays inherent in studying a chronic disease with survival time as a primary outcome of concern.
Second, AIDS is a new disease. Our understanding of the natural history is growing rapidly, but is still very limited compared to diseases that we have been studying for decades or even centuries. The disease is complicated, with large numbers of potentially fatal sequelae, each of which may require a different therapeutic and/or prophylactic approach. Each new effective regimen itself alters the "natural history"; our success in reducing the incidence of any of these infections results in an increased incidence of other opportunistic diseases.
Finally, the nature of the populations that have borne the brunt of the epidemic is unusual. At least in the United States, AIDS is a disease of homosexuals, of intravenous drug users, of minorities, and of the poor--all groups that are stigmatized to varying degrees by "mainstream" society. Because some of these groups are virtually removed from the health care system, it is difficult to involve them in studies of investigational drugs. The gay population, on the other hand, had become highly politicized over civil rights issues even before AIDS; the advent of the disease has heightened their concerns substantially. Representatives of affected communities have become increasingly vocal regarding issues related to testing of investigational drugs (Harrington 1990; Ellenberg et al., in press).
All these issues generate intense and often conflicting concerns for statisticians designing clinical trials to evaluate AIDS therapies. The development of more efficient clinical trials designs long has been a research topic of interest to statisticians, and the impact of the pertinent issues on trial designs have been discussed by Byar et al. …