We share Dr. Stolley's concerns about the inadequacy of nonrandomized studies to identify and evaluate reliably the effects of experimental drugs. Our experience, however, is entirely inconsistent with his premise that there exists a "trend toward abandonment of randomization in large community-based trials" [1].

The Terry Beirn Community Programs for Clinical Research on AIDS (CPCRA), sponsored by the National Institute of Allergy and Infectious Diseases, was developed specifically to afford community practitioners the opportunity to participate in randomized trials. This program, initiated in 1989, has been extremely successful; 20 studies are in progress or have been completed in community settings; only 4 of the studies were nonrandomized because of the observational nature of the study design. Some of these randomized trials were planned to enroll upward of a 1000 patients; one about-to-be-initiated trial comparing the strategies of early and late combination antiretroviral therapy may be the largest trial of therapy for human immunodeficiency virus infection ever conducted. Most of these trials have been received with great enthusiasm; many have completed enrollment more rapidly than was projected.

Our observation is that treatment research in the community setting is of an increasingly randomized nature, contrary to Dr. Stolley's assertion. The ComPACT 1 study of immediate and delayed antiretroviral therapy, sponsored by the Community Consortium in the San Francisco Bay Area, is another large randomized trial, and it is assessing a question that is of heightened importance in light of the recent Concorde study in Europe [2]. The expanded access programs developed by pharmaceutical companies to run in parallel with more rigorous trials are being implemented as randomized dose-comparison studies [3].

Patients and community activists are highly knowledgeable about the state of the art of therapy for HIV infection and AIDS, and they are more sophisticated about clinical research methods (and particularly clinical trial design) than many physicians. They increasingly appreciate the need for unbiased assessments of treatment efficacy and for pragmatic but reliable evaluation of optimal treatment strategies, and they well understand that compassion and scientific rigor are not inherently incompatible [4].

Editors' Note—Dr. Stolley chose not to respond.