pare the characteristics of a previous exposure to NSE. The Committee identified 26 prospective cohort studies, which enroll participants based on their risk characteristics, and follow them to compare related outcomes. The committee felt that 14 of these studies were especially strong in terms of study design and relevance (and noted those studies with an asterisk in a table in Appendix D). Case-control and prospective cohort studies were ranked as having the strongest available study design.
The Committee also identified six ecological studies, which examine populations rather than individuals and cannot establish causal links. Finally, the Committee identified many cross-sectional and serial cross-sectional studies. Cross-sectional studies describe the associations between a disease and risk factors in a population at a specific point in time. The Committee considered such studies as having the weakest design because causal inferences cannot be drawn from them. Serial cross-sectional studies examine groups of people at multiple time points, and offer stronger evidence of shifts in associations over time. As opposed to prospective cohort studies which examine individual-level changes in risk behavior, well-designed serial cross-sectional studies can indicate patterns of behavior change at the community level. As supporting evidence, the Committee included six cross-sectional and four serial cross-sectional studies in Appendix D, based on their strong study design and relevance to the Committee’s statement of task.
The Committee used caution in interpreting the results of studies reviewed in this chapter because of their generally weak designs and serious limitations. One limitation is that the studies identified do not randomly assign subjects to treatment and control groups—rather, participants deliberately choose whether to use NSEs and other services. This creates an unavoidable risk of selection bias, and means that differences in rates of risk behaviors and HIV infection may not be due to use of the service itself. Another limitation is that the study designs generally do not allow separate examination of program elements, so the independent contribution of improving access to sterile needles and syringes cannot be assessed. For example, NSE is often one component of a multi-component HIV prevention program, making it difficult to isolate the exact effects of NSE alone.
Another concern is that studies of drug abuse, like most behavioral research, depend heavily on self-reported data on drug use, risk behavior, and precautions taken to reduce risk. Studies evaluating the effectiveness of NSEs are no exception. Self-reported data can introduce bias, as drug abuse is illegal in most settings, and drug users may underestimate risk behavior and overestimate protective behavior. Still, the self-reports of drug users on the incidence of drug abuse and drug-related risks have generally been shown to be valid (Darke, 1998) and remain the major type of outcome measures used in studies of NSE.