To determine the impact of the implementation of a needle-exchange program (NEP) on the spread of human immunodeficiency virus (HIV) in an injection drug user (IDU) community. We conducted a Monte Carlo simulation study of a theoretical population of 10000 IDUs. The population was followed monthly from 1984 to 2000. HIV was assumed to be transmitted only by needle sharing. The NEP was introduced in 1989 and evaluated over a period of 11 years. The impacts of the proportion of the population attending the NEP, the risk level of IDUs attending the NEP, the reduction in needle-sharing frequency, and the number of new needle-sharing partners acquired at the NEP on prevalence and incidence of HIV were determined. Increasing the proportion of the population who always attend the NEP and eliminating needle-sharing incidents among IDUs who always attended the NEP were the most effective ways of reducing the spread of HIV. Attracting high-risk users instead of lower risk users to the NEP also reduced the spread of HIV, but to a lesser extent. NEPs are effective at reducing the spread of HIV; even under the worst case scenario of low risk users more likely to attend the NEP, one additional partner per month as a result of attending the NEP, and poor NEP attendance, the estimated prevalence was still less than that from the scenario without an NEP. Under our model, NEPs were shown to reduce the spread of HIV significantly. Efforts should be focused on getting as many IDUs as possible to become regular NEP attenders and stop sharing needles rather than partially reducing the frequency of sharing by a larger number of IDUs.