potential rival explanation still would not explain why no such abrupt decline was observed with the sexually transmitted cases. Moreover, as stated earlier, Pierce County has been part of CDC's hepatitis surveillance system since the late 1970s, and no changes in protocol have been implemented since then. Furthermore, no hepatitis vaccine campaign was ever undertaken in the county (H. Hagan, personal communication, January 1995).
Social desirability effects (telling the interviewer what you believe he or she wants to hear) might explain why needle exchange users report large reductions in their risk behavior and lower frequencies of injection risk behavior than injection drug users not using the exchange (even though the interviews were conducted by research staff separate from the exchange staff). The alternative explanation of social desirability for self-reported risk behavior does not match the findings from the hepatitis case-control studies (Hagan et al., in press), in which injection drug users who used the exchange were much less likely to have been recently infected with hepatitis B and hepatitis C viruses.
Self-selection or volunteer bias might explain why the injection drug users who use the exchange report low levels of current risk behavior and have low incidence of HIV, hepatitis B, and hepatitis C virus infections. The exchange might simply have attracted injection drug users who were concerned about health and were practicing safer injection prior to the implementation of the needle exchange. However, this alternative explanation is not consistent with the exchange users' report of their frequent risk behavior prior to beginning use of the exchange and the community-wide reduction in hepatitis B incidence and the low community prevalence rate of HIV infection.
Case-control studies are always open to the objection that cases and controls may differ in important respects. If cases and controls in the Tacoma study (Hagan et al., in press) were not arising from the same source population, it would result in biased estimates of the associations between needle exchange participation and incident cases of hepatitis B and C. Nonequivalence between controls and cases was not detected on measured demographic characteristics for hepatitis C (sex, race, age, and duration of injection). Hepatitis B cases and controls differed regarding age and duration of injection. Nonetheless, after statistically controlling for such disparities, the association between needle exchange participation and hepatitis B was still found to be significant.
Another concern that may arise is that hepatitis B cases may not be representative of nonselected cases in the community because they were symptomatic. It is true that immunocompromised individuals (e.g., from renal dialysis, steroid use, HIV, and age extremes) are less likely to be symptomatic. However, given the low HIV prevalence, the age range of