The study described here presents an innovative approach to analyzing intervention outcomes among women substance abusers participating in a national HIV prevention research study funded by the National Institute on Drug Abuse. We used cluster analysis to divide the women in our sample (N=557) into four distinct subgroups predominantly characterized by differences in drug use, injecting risk, sexual behaviors, and drug and sexual risk combined. The four subgroups resulting from this process were primary crack-using women, primary needle-using women, high-frequency needle-using women, and women with multiple drug and sex risk behaviors. Our analysis focuses on changes in self-reported risk behaviors from baseline to 6-month follow-up. In general, the results clearly indicate that the women are heterogeneous; that is, the subgroups exhibit varying patterns of drug use, injecting risk, sexual behavior, and HIV seropositivity. Significant outcomes were found in many areas, indicating positive changes in risk behaviors. The two smaller subgroups of women--high-frequency needle users and those in the multiple-risk behavior subgroup--reported the highest rate of high-risk behaviors and seropositivity but also showed the greatest change at follow-up. A particularly important finding resulting from our analytical approach is that well over half the women in our sample were primary crack users (n=313). This finding is even more significant in light of the fact that the Cooperative Agreement specifically tried to include 70% or more participants who were injectors. Although the rate of HIV seropositivity is not as high for this crack-using subgroup as for the two smaller needle-using subgroups, a greater number of "women who are HIV positive" are in this primary crack-using subgroup than in all the other subgroups. Most of the crack-using women reported that they were not currently injecting drugs and never shared needles, but 10% were seropositive for HIV, suggesting that their risk comes primarily from sexual behaviors. Behaviors in this larger subgroup of women did not change as dramatically as those of women in the smaller subgroups; however, the women did show improvement in areas related to indirect risk (e.g., alcohol and crack use) and in several areas where change is most needed (e.g., trading sex for drugs and using condoms). The results demonstrate a promising alternative approach to analyzing substance abuse and HIV risk behaviors, and they suggest the need for further research on alternative interventions for women with different patterns of risk behaviors.