sharing cotton, cookers, or water, and backloading (see Wiebel, 1990, for a description of the measure). Two additional dichotomous drug-related variables were constructed, one based exclusively on whether the injection drug user used other injection drug users' needles without using bleach and the other based on whether drug users passed their own used needles to other injectors. A final syringe-sharing variable consistent with Watters et al. (1994)—injecting risk behavior—was derived. This latter variable is based on information concerning an injector's use of someone else's needle and ignores possible cleaning practice with bleach. The sexual risk behavior variable assessed whether a respondent had multiple sex partners in the past 6 months or an injection drug user sex partner and did not always use a condom.
Three measures of needle exchange were derived: dichotomous use/nonuse in the past 4 weeks; use over two interview waves; and a four-category measure of frequency of use similar to the Watters et al. (1994) measure. Other information collected during the follow-up and baseline interviews included: number of injection drug users' associates and cohabitants; frequency of worry about becoming infected with HIV via injection drug use; and having enough money.
Of the overall 1994 data (n = 728), 8 percent used the needle exchange once in the previous 4 weeks; 12 percent used it 2 to 3 times; and 19 percent used it once a week or more. This represents 40 percent (n = 285) of the total sample having used the exchange in the past 4 weeks, which also represents approximately 21 percent of all injection drug users who had used the program at least once since February 1, 1994.
With respect to HIV risk behaviors assessed in 1994, a comparison of needle exchange users and nonusers revealed no statistically significant difference in the proportion of study participants reporting having engaged in injection drug use risk behaviors, whereas nonusers were found to be more likely to report multiple sex risk (43 compared with 40 percent). The investigators used logistic regressions to examine the effect of exchange use on each risk behavior when prior level of risk was accounted for (using 1993 follow-up data on the 1988 and 1992 cohorts; n = 405). The results showed that the measure of prior risk behavior was the only significant predictor of current risk level; the main effect of needle exchange use, as well as its interaction with age, was nonsignificant for all risk variables. Moreover, these logistic regressions were reexamined using the 1994 data as a cross section, ignoring prior risk behavior, and the results showed that each risk variable was unaffected by frequency of exchange use.
There were 258 current injection drug users from the 1988 and 1992 cohorts known to be at risk of seroconverting at the 1994 follow-up. An equal proportion of these HIV-negative study participants were needle exchange users and nonusers. Three seroconverted over an average 9-month