randomized. Moreover, given the sensitive nature of these programs (in contrast with smoking cessation programs, for example), many volunteer communities may not be capable of initiating proper legislative change (of paraphernalia and prescription laws) to legally allow such programs to take place. Furthermore, given the results of the two recent government-sponsored reports that have concluded that these programs have positive effects and do not appear to have negative impacts (U.S. General Accounting Office, 1993; Lurie et al., 1993), it may not be ethical to withhold treatment from communities willing to initiate such programs. Problems may also arise because communities eager to participate may proceed with program implementation after having been informed that they have been assigned to the control condition.
In addition, one major concern with these designs in prevention research is differential attrition rates across conditions. As Booth and Watters (1994) point out in their review of risk reduction interventions, participation in the treatment condition is more demanding than in the control condition, which typically leads to experimental attrition.
Cost is a factor, particularly when treatments are randomized across large units, such as cities or communities, rather than across individuals. Only a small number of units may be assigned to conditions. Yet the strength of randomization depends on the random assignment of a sufficiently large number of units to substantially weaken the possibility that confounding factors would coincidentally vary with the random assignment. When only a small number of units are randomly assigned to conditions, given the low probability that the treatment and control conditions will be equivalent on unmeasured factors, sensitivity is necessarily reduced.
The panel recommends adopting strong observational epidemiologic designs (i.e., prospective and case-control studies) rather than attempting to conduct large-scale randomized experiments to evaluate needle exchange and bleach distribution programs.
To better understand the workings of needle exchange and bleach distribution programs and how to render them more effective, we need a deeper understanding of many phenomena that are not specific to needle exchange and bleach distribution programs but are more general in scope. For example, we need to know about the processes of addiction, about the propagation of infectious diseases, about the dynamics of social networks, about the underlying factors in personal failure and success, and about the role of sexual behaviors in the lives of injection drug users.