We sought to validate an in vitro system which could predict the minimal effect dose of antiretroviral agents. Mixtures of uninfected CEM cells and CEM cells chronically infected with human immunodeficiency virus (HIV) type 1 MN were exposed to 2',3'-didehydro-3'-deoxythymidine (D4T) in vitro in a hollow-fiber model which simulates the plasma concentration-time profile of D4T in patients. Drug concentration was adjusted to simulate continuous intravenous infusion, or an intravenous bolus administered twice daily. The effect of the dosing regimen was measured with viral infectivity, p24 antigen, and reverse transcriptase or PCR for unintegrated HIV DNA. Dose deescalation studies on a twice-daily dosing schedule predicted a minimum effect dose of 0.5 mg/kg of body weight per day which correlated with the results of a clinical trial. Antiviral effect was demonstrated to be independent of schedule for every 12-h dosing versus continuous infusion. Finally, at or near the minimal effect dose, efficacy appeared to depend on the viral load. The ability of this in vitro pharmacodynamic model to assess the response of HIV-infected cells to different doses and schedules of antiviral agents may be useful in the design of optimal dosing regimens for clinical trials but requires validation with other types of antiretroviral agents.