A mathematical model of influenza transmission dynamics is used to simulate the impact of neuraminidase inhibitor therapy on infection rates and transmission of drug-resistant viral strains. The model incorporates population age structure, seasonal transmission, immunity and inclusion of elderly nursing home residents or non-residents. Key parameter values are estimated from epidemiological, clinical and experimental data. The analysis examines the factors determining the population spread of antiviral resistance, and predicts no significant transmission of neuraminidase inhibitor resistant virus. This conclusion is robust even at high therapy levels and under conservative assumptions regarding the likely frequency of transmission of resistant virus. The predicted incidence of resistance following protracted usage reflects primary drug resistance, currently estimated as approximately 2% for neuraminidase inhibitor therapy. It is also shown that until high levels of therapy are attained, early treatment of symptomatic cases is more efficient (per unit of drug) at preventing infections than prophylactic therapy.