The risk of developing active tuberculosis is highest within the first 2 years of infection. Therefore, an intervention that targets persons with recent infection, such as identifying contacts of active cases, could be particularly effective as an epidemic control measure. A mathematical model of a tuberculosis epidemic is formulated and used to evaluate the strategy of targeting therapy to persons with recently acquired latent tuberculosis infection. The model is used to quantify the effectiveness of therapy for early latent tuberculosis infection in reducing the prevalence of active tuberculosis. The model is also used to demonstrate how effective therapy for early latent tuberculosis infection has to be to eliminate tuberculosis, when used in conjunction with therapy for active tuberculosis. Analysis of the model suggests that programs such as contact investigations, which identify and treat persons recently infected with Mycobacterium tuberculosis, may have a substantial effect on controlling tuberculosis epidemics.