Aims To study the effectiveness of an educational intervention on risks associated with drug injection, comparing primary [unsafe HIV-hepatitis C virus (HCV) practices] and secondary (local complications at injecting site) end-points in harm reduction (HR) programmes offering this intervention versus HR programmes not offering it. Design This non-random clustered intervention study was conducted in nine intervention groups (programmes offering the intervention) and eight control groups (programmes not offering it). Each participant was followed-up through a telephone interview at enrolment and at 6 and 12 months. Setting The study took place in 17 cities throughout France. Participants Of the 271 participants, 144 were enrolled into the intervention group and 127 in the control group. Of the latter, 113 received at least one educational session. Intervention A series of participant-centred face-to-face educational sessions. Each session included direct observation by trained non-governmental organization (NGO) staff or volunteers of participants' self-injecting the psychoactive product they used habitually; analysis by the trained NGO staff or volunteers of the participant's injecting practices, identification of injection-related risks and explanation of safer injecting practices; and an educational exchange on the individual participant's injection practices and the questions he or she asked. Measurements Primary and secondary outcomes were 'at least one unsafe HIV-HCV practice' and at least one injection-related complication (derived from a checklist). Findings The proportion of participants with at least one unsafe HIV-HCV practice in the intervention group decreased significantly, from 44% at M0 to 25% at M6, as well as complications at the injection site (from 66 to 39% at M12), while in the control group it remained mainly stable. Multivariate probit analyses showed that the intervention group experienced a significant reduction in unsafe HIV-HCV practices at M6 [coefficient, 95% confidence interval (CI) = À0.73 (À1.47 to 0.01)] and in injection-related complications at M12 [coefficient, 95% CI = À1.01 (À1.77 to À0.24)], compared with the control group. Conclusions An inexpensive and easily implemented educational intervention on risks associated with drug injection reduces significantly unsafe HIV-HCV transmission practices and injection-related complications.