The effect of an open arm in the enrollment to a randomized clinical trial comparing zidovudine (ZDV) plus didanosine (ddI) versus ZDV plus zalcitabine (ddC) was assessed. HIV-infected individuals were eligible to participate in this protocol if they were ddI and ddC naive, had CD4 counts of 50-350/mm3, and were residents of the province of British Columbia. Participating individuals could choose between open-label ZDV/ddI, ZDV/ddC, or randomization to open-label ZDV/ddI or ZDV/ ddC. Study drugs were made available free of charge for all participants through a centralized drug distribution system. There is no other source of these drugs in the province. Primary care physicians were required to renew the patient's prescription every 2 months. Enrollment was initiated in November 1992 and was closed in March 1994 when the randomized arm of the protocol met the predetermined target sample size of 120 evaluable participants. A total of 582 patients received combination therapy in the province through this protocol: 138 (28%) enrolled in the randomized arm and 444 (76%) in the open arm. In the latter group, 320 (72%) were initially prescribed ZDV/ddI and 124 (28%) were prescribed ZDV/ddC. The enrollment rate was strikingly higher in the open arm, with 168 patients enrolled in the first 2 months compared with 138 patients enrolled in the randomized arm over 17 months. Of the 78 study physicians, 69 enrolled patients in the open arm and 23 enrolled patients in the randomized arm of the study. Experienced physicians were more likely to refer patients for randomization (p = 0.025). No statistically significant differences were observed between patients enrolled in either study arm. Our results illustrate the challenge posed to recruitment into clinical trials by the coexistence of an open arm. This is despite the noncoercive, open-label and community-based nature of our randomized protocol and the high priority given to it by a variety of local and national organizations. It is clear that an increased commitment by all interested parties will be required if randomized clinical trials are to be carried out with coexistent open arms.