A cross-sectional study was performed to explore the behaviour of female intravenous drug users (lOUs) in Madrid, Spain, to determine the relationships between behaviour and a number of dependent variables. We considered: l.Sociodemographic characteristics Drug-taking behaviour, sexual behavior (including prostitution), reproductive health, and utilisation of services targeted at female IDUs. 2. Direct and indirect determinants of HIV in the study population 3. Determinants of potential risk factors for HlV 4. Attitudes among women towards safer sexual practices 5. The effects of different kinds of drug treatment on risk behavior The limitations inherent in a cross-sectional study design, such as the time sequence of events, were addressed appropriately in the questionnaire. Because of the potential for selection bias, we implemented sampling techniques (such as sampling from multiple sites) to correct for this. Finally, in order to explore at different stages how the different variables might have a direct and indirect effect on the dependent variables and how the effect may be mediated through other variables in the causal pathway, we built a number of hierarchical models. The foremost conclusion that can be drawn is the significance of low socioeconomic status. Furthermore, the high-risk population is homogenous and represents an already marginalised population. The result of regression analysis showed the most important and most predictive variable (OR=12.2) to be having a regular partner who is HIV-positive. Other significant factors were: age 35 or over; no fixed address; history of sharing needles; and co-infection with an STD. Early age of first pregnancy (younger than 18) was a direct determinant of prostitution and needle sharing. We also detected a link between sexual risk behaviour and drug-taking risk behaviour. The public health implications of these findings seems obvious. Women have a range of special needs, including reproductive needs. Efforts directed towards peer education for men and women are likely to be successful since the person delivering the intervention is someone the target population can identify with.