In this paper, the authors propose a method for estimating the incidence of heroin use by adjusting reported numbers of heroin users visiting drug treatment agencies for the time lag between onset of heroin use and first treatment request (lag distribution). The adjusted incidence is relative, since it represents the number of individuals beginning heroin use in each year whose cases will be reported within 8 years of starting use. Users with longer lag times or whose cases are never reported are excluded. Utilizing data from southeastern England (1991--1998), the authors analyzed the effects of covariates (sex, age group, ethnic group, route of consumption, and year of onset of drug use) on the lag distribution. Trends in the adjusted incidence of heroin use were very different for injectors and noninjectors: Incidence among injectors seemed to be stable, while in noninjectors it increased twofold between 1991 and 1996--1997. These results must be interpreted cautiously, especially in relation to the wider context of underlying trends in the population. Potential biases derive from underreporting and from changes in the proportion of heroin users in treatment. The lag correction method adds substantially to the value of routine treatment data, at least for heroin use, and is potentially the best method for obtaining estimates of incidence.