and they can mask significant variations and changes in small local population subgroups. However, the Drucker and Vermund model allows the use of local serosurvey data among various injection drug user subgroups to estimate seroprevalence among key population subgroups in a given geographic area.
Despite the wide variation by geographic location, evidence suggests that, even in areas in which prevalence is low among injection drug users, the risk of HIV infection should not be viewed with complacency. In Milan, Edinburgh, New York City, and Bangkok, once HIV became established in a community of injection drug users (i.e., a prevalence of less than 10 percent), prevalence subsequently rose dramatically within the next 2 to 4 years (Angarano et al., 1985; Robertson et al., 1986; Des Jarlais and Friedman, 1988a; Des Jarlais et al., 1994; Kitayaporn et al., 1994). Although this pattern has not been observed universally (some cities, such as Los Angeles, stabilize prevalence at lower levels), these examples suggest the need for HIV prevention programs not only when prevalence is moderate or high, but also when prevalence is low (especially when injection drug users frequently engage in high-risk behaviors). It is interesting to note, in a study of four cities with sustained low seroprevalence of HIV in injection drug users, all cities had extensive HIV prevention efforts (Des Jarlais, 1994).
Hahn and colleagues' review of injection drug users in treatment (Hahn et al., 1989) reported that the HIV seroincidence data available from some sites revealed rates of 0 to 14 percent per year, again, with the highest incidence reported in the Northeast, especially in New York City. Similar trends were observed among surveys conducted nationwide (again, in a variety of treatment settings) by CDC (Allen et al., 1992). Here, annual seroincidence rates ranged from 0 to 43 percent, the highest rates (ranging from 15 to 44 percent) again reported in the Northeast.
One study provided prevalence and incidence rates for HIV infection among intravenous drug users in and out of treatment in Philadelphia (Metzger et al., 1993). In 1989, prevalence was 10 percent among intravenous drug users in methadone treatment and 16 percent among those out of treatment. Eighteen months later, follow-up rates revealed incidence rates of 3.5 percent for those intravenous drug users in treatment, and 22 percent for those out of treatment. These data, although subject to possible selection bias, suggest that treatment for drug abuse is probably an important intervention for HIV prevention.
Efficiency of transmission among injection drug users varies according to many behavioral risk characteristics, which vary among individuals. They include: frequency of injection, sharing injection equipment, the number of needle-sharing partners, and risky sexual practices (discussed below). However, in most studies that have examined risk factors and seroconversion or prevalent