Volunteer bias in nonrandomized evaluations of the efficacy of needle-exchange programs

Objective. Nonrandomized comparisons of the incidence of HIV and hepatitis B and C between injection drug users (IDUs) who do and do not attend volunt ary needle-exchange programs may be subject to bias. To explore possible so urces of bias, we examined characteristics associated with voluntarily begi nning or ceasing to participate in the Seattle needle exchange. Methods. In a cohort of 2,879 IDUs, a standardized questionnaire measured c haracteristics present at enrollment. We examined the relation of these cha racteristics to the proportion of IDUs who began to use the program during the ensuing 12-month follow-up period and to the proportion of current exch angers who dropped out during that period of time. Results. Of the 494 never-exchangers at baseline, 32% attended the exchange program during follow-up; those who reported sharing syringes or who were homeless at enrollment were more likely to become new exchange users (adjus ted risk ratio [ARR] for becoming an exchange user = 1.8 for those who shar ed syringes, and ARR = 2.2 for those who were homeless). Of 1,274 current e xchangers, 16% stopped using the exchange during followup, with daily injec tors (ARR = 0.6) and those who reported backloading (ARR = 0.6) being relat ively less likely to drop out of the exchange. Conclusions. The analysis suggests that IDUs participating in needle-exchan ge programs ata given point in time may include a particularly high proport ion of those injectors whose pattern of drug use puts them at elevated risk of blood-borne viral infections.