failed to show any statistically significant trends. (More sophisticated, multivariate models were also used, with similar results.)
Although drug behaviors and gender remained statistically constant over time, Kaplan and colleagues (Heimer et al., 1993; Kaplan, 1994a) reported that the racial composition of the study population did change over time; the percentage of white participants steadily increased (from 17 to 39 percent). Because white participants are less at risk, the change in HIV prevalence might be due to the change in the composition of the population served, but, as the authors note, this does not appear to be a viable plausible explanation for the decline in the infectivity of needles. The decline occurred in the first 120 days and then stabilized. The number of white participants steadily increased throughout the duration of the project. The magnitude of the change in composition is also not great enough to explain the reduction in HIV seropositivity. Specifically, the authors state (Heimer et al. 1993:219):5
At the start of the program, 83% of the participants were nonwhite and 67.5% of the syringes were infected. After 7 months, 62% of the clients were nonwhite and 43% of the syringes were infected. If needle exchange were without effect, then 87.3% of the syringes returned by nonwhites ought to test positive, a plausible result. However, -29.2% of the syringes returned by whites would need to test positive, clearly an impossible result.
Also, Heimer et al. (1993) argue that the decrease in prevalence was not the result of reduced needle sharing, because the fraction of syringes assigned to one person but returned by someone else did not change. As noted by the authors, because of the manner in which the exchange operated, it is not possible to know with certainty whether discordant needles (i.e., exchanged by one individual and returned by a different person) had been shared or whether they were simply exchanged by someone else.
Finally, Kaplan and others (Kaplan, 1994a; O'Keefe et al., 1991) report that up to 60 percent of the New Haven needle exchange program participants dropped out. Although the status of those who dropped out was not always clear, as noted above, a small fraction of these participants were known to have entered drug treatment. Kaplan (1994a) argues that even short-term exposure to the needle exchange program could contribute to its aggregate impact.
The pattern of evidence surrounding the New Haven needle exchange program involves a set of models, driven in large measure by empirical data gathered from participants and the needles they exchanged. Although the