Statistical issues in the design of HIV vaccine trials.

HIV vaccine trials present significant challenges related to trial endpoints, vaccine efficacy measurement, and the role of nonvaccine interventions. Infection is a valid endpoint for detecting sterilizing immunity. But if the vaccine prevents AIDS without preventing infection, infection may be a misleading surrogate. Appropriate endpoints must be defined for other mechanisms of vaccine action. Direct, indirect, behavioral, and biological effects all determine vaccine efficacy. False security among HIV-vaccine recipients may make negative behavioral effects an important component of vaccine performance. Both biological potency and a more comprehensive program effectiveness should be measured. These goals may require unblinded designs or community randomization. Nonvaccine interventions are currently the only HIV-prevention strategy. Support for larger scale implementation requires more rigorous evaluation that is less dependent on self-reported behavioral changes. The vaccine trial cohorts provide a unique opportunity to cost-effectively evaluate behavioral interventions.