programs has a substantial protective effect for preventing new HIV infections. However, the results need to be interpreted with care. That is, nonequivalence across groups being compared (needle exchange users versus nonusers) precludes making strong causal inferences about the direct effect of the needle exchange on HIV incidence rates. Nonetheless, these data do reflect a significant association between needle exchange participation and HIV infection (Des Jarlais et al., 1994b, 1995).
The most recent studies that have examined drug-use behaviors among needle exchange participants show either stable levels of reported drug injection frequency or even slight declines over time among injection drug users who continue to participate in needle exchange programs (Watters et al., 1994; Paone et al., 1994a; Des Jarlais et al., 1994a; Oliver et al., 1994). In the recent New York City study, Paone et al. (1994a) reported a statistically significant decrease in injection frequency among needle exchange participants, from 95 times in the 30 days preceding program participation to 86 times in the 30 days prior to participants' being interviewed.
The only exception to this reported trend comes from an unpublished research manuscript from Chicago researchers (O'Brien et al., 1995a). As noted in the Preface, as the panel was concluding its deliberations, the Assistant Secretary for Health made public statements that a number of unpublished needle exchange evaluation reports had raised doubts in his mind about the effectiveness of these programs. The panel deemed these statements to be significant in the public debate, therefore necessitating appropriate consideration in order for the panel to be fully responsive to its charge. The panel therefore reviewed the unpublished studies, one of which was the aforementioned O'Brien et al. (1995a) Chicago study. As unpublished findings, this research lacks the authority provided by the peer review and publication process. For this reason, the panel gave special attention to scrutinizing and describing in detail results reported by the researchers, as well as appraising their probative value (see Appendix A).
The investigators infer from their findings that those who participate in needle exchange programs spend more money and inject more frequently than nonparticipants as a result of their participation in the program. Their assertion is based on data that, according to these authors, support the contention that program participation is economically driven (i.e., by the cost of needles). The panel's review raised serious concerns about the tenability of their inferences. For instance, a clearly insufficient theoretical and empirical development of the underlying models is used. That is, there are