In vitro selection studies and data from large genotype databases from clinical studies have demonstrated that tenofovir disoproxil fumarate and abacavir sulfate select for the K65R mutation in the human immunodeficiency virus type 1 polymerase region. Furthermore, other novel non-thymine nucleoside reverse transcriptase (RT) inhibitors also select for this mutation in vitro. Studies performed in vitro and in humans suggest that viruses containing the K65R mutation remained susceptible to zidovudine (ZDV) and other thymine nucleoside antiretroviral agents. Therefore, ZDV could be coformulated with these agents as a "resistance repellent" agent for the K65R mutation. The approved ZDV oral dose is 300 mg twice a day (b.i.d.) and is commonly associated with bone marrow toxicity thought to be secondary to ZDV-5'-monophosphate (ZDV-MP) accumulation. A simulation study was performed in silico to optimize the ZDV dose for b.i.d. administration with K65R-selecting antiretroviral agents in virtual subjects using the population pharmacokinetic and cellular enzyme kinetic parameters of ZDV. These simulations predicted that a reduction in the ZDV dose from 300 to 200 mg b.i.d. should produce similar amounts of ZDV-5'-triphosphate (ZDV-TP) associated with antiviral efficacy (>97% overlap) and reduced plasma ZDV and cellular amounts of ZDV-MP associated with toxicity. The simulations also predicted reduced peak and trough amounts of cellular ZDV-TP after treatment with 600 mg ZDV once a day (q.d.) rather than 300 or 200 mg ZDV b.i.d., indicating that q.d. dosing with ZDV should be avoided. These in silico predictions suggest that 200 mg ZDV b.i.d. is an efficacious and safe dose that could delay the emergence of the K65R mutation.