To provide time-trend estimates of HIV-1 prevalence among injecting drug users (IDU) in London. HIV-1 prevalence and HIV testing behaviour were measured in four serial point prevalence surveys of IDU recruited at multiple sites in community-based non-treatment and drug treatment settings between 1990 and 1993. Community subjects were recruited through social network sampling by trained indigenous interviewers; treatment subjects were interviewed at agencies. With informed consent, subjects responded to a structured questionnaire covering risk behaviour. Volunteered saliva samples were tested anonymously for anti-HIV-1. Statistical comparisons across years were examined using mixed binomial logistic and log-linear models. Pearson's chi 2 and Fisher's exact tests were also used for some two-group comparisons. Similar samples were recruited each year. HIV-1 prevalence rate declined from 12.8% in 1990, 9.8% in 1991, 7.0% in 1992, to 6.9% in 1993. The statistical modelling suggested that the overall trend in prevalence rates was one of decelerating decline. There was no difference in prevalence rate by gender and length of injecting. Less than one-half (46%) had received a named HIV test. Over one-half of the HIV-positive IDU were unaware of their status. The pattern of decline in HIV prevalence rate is attributed to changes in risk behaviour following HIV prevention interventions.