Plasma didanosine concentration data from 36 patients receiving once-a-day therapy and from 33 patients receiving twice-a-day therapy were subject to population pharmacokinetic analysis with the computer program NONMEM. Once- or twice-a-day regimens of didanosine were administered intravenously (i.v.) (dose: 0.8-33 mg/kg) during the first 2 weeks of therapy, and orally (dose: 1.6-66 mg/kg) for the remaining 4 weeks of therapy. Plasma pharmacokinetics were determined after the first and last (steady-state) i.v. and oral doses. Population pharmacokinetic parameters for the combined i.v. and oral steady-state data were (mean [%CV]): systemic clearance, CL, 0.70 (5.2) L/h/kg; central compartment volume, Vc, 0.18 (32) L/kg; steady-state distribution volume, Vdss, 0.84 (6.8) L/kg; first-order absorption rate constant, Ka, 1.3 (9.5) hr-1; and bioavailable fraction, F, 0.34 (8.5). Interindividual variability (omega) was (%CV) 22.3 and 71.0 for CL and Vc, respectively. Intraindividual (residual) variability (sigma) in plasma concentrations (%CV) was 50.2. Body weight, sex, and age did not account for the variability in either CL or Vc, and the use of alternate pharmacokinetic models did not reduce the value of intraindividual variability. Population parameters for the combined i.v. and oral first-dose data were generally similar to those for the steady-state data. The parameters can be used to design dosing regimens in patients using the Bayesian feedback approach.